Morphologic features of neuroblastoma (Schwannian stroma-poor tumors) in clinically favorable and unfavorable groups

Authors


  • The meeting of the INPC members in Vienna, Austria, was sponsored by the companies Zeiss, Olympus, and Oncor.

  • Participating principal pathologists and pediatric oncologists: Department of Clinical Pathology, University of Vienna, Vienna, Austria (G. Amann, M.D.); St. Anna Children's Hospital, Vienna, Austria (R. Ladenstein, M.D. and H. Gadner, M.D.); Department of Pathology, University of Graz, Graz, Austria (E. Spuller, M.D. and M. Ratschek, M.D.); Department of Pediatrics, University of Graz, Graz, Austria (R. Kerbl, M.D.); Department of Pathology, University of Innsbruck, Innsbruck, Austria (A. Kreczy, M.D.); Institute of Pathological Anatomy, University of Salzburg, Salzburg, Austria (O. Dietze, M.D.); Hospital of Klagenfurt, Klagenfurt, Austria (W. Pobegen, M.D.); Hospital of Leoben, Leoben, Austria (I. Mutz, M.D.); Department of Pediatric Oncology, University of Gent, Gent, Belgium (G. Laureys, M.D.); Department of Pathology, University of Gent, Gent, Belgium (C. R. De Potter, M.D.); Department of Pediatric Pathology, University of Kiel, Kiel, Germany (D. Harms, M.D.); Children's Hospital, University of Cologne, Cologne, Germany (B. Hero, M.D. and F. Berthold, M.D.); Institute of Pathology, University of Tübingen, Tübingen, Germany (E. Kaiserling, M.D.); Department of Oncology and Hematology, Olgahospital Stuttgart, Stuttgart, Germany (F. Schilling, M.D.); Children's University Hospital, University of Tübingen, Tübingen, Germany (R. Handgretinger, M.D.); Institute of Pathology, University of Hamburg, Hamburg, Germany (R. Schäfer, M.D.); Department of Pathology, Giannina Gaslini Children's Hospital Genoa, Genoa, Italy (C. Gambini, M.D.); Hematology-Oncology Department, Giannina Gaslini Children's Hospital Genoa, Genoa, Italy (B. De Bernardi, M.D.).

Abstract

BACKGROUND

After the establishment of the International Neuroblastoma Pathology Classification system, the authors studied retrospectively the prognostic impact of morphologic features in a series of two clinically distinct subsets of patients with peripheral neuroblastic tumors (NTs), i.e., tumors in the neuroblastoma category.

METHODS

Forty-seven NTs categorized into either clinically favorable or unfavorable subgroups were selected randomly from 100 NTs for a histologic review that included the evaluation of 14 morphologic characteristics. The review was performed individually followed by a group review. The correlations of the prognostic significance of the individual morphologic features and the correlations among them were determined by use of odds ratios (ORs) with corresponding 95% confidence intervals (95%CIs). The inter-rater agreement was determined by using the Cohen κ coefficient.

RESULTS

Ten of 14 morphologic features, including nuclear size, cellularity, prominent nucleoli in undifferentiated or poorly differentiated neuroblasts, and the number of mitotic and karyorrhectic cells (MKI), showed a significant correlation with the clinical groups (ORs between 36.9 and 10.5 and P values between < 0.001 and 0.002). In addition to the patient's age at diagnosis (OR, 7.4; 95%CI, 1.9–28.9; P = 0.002), 8 of 14 features also provided prognostic information (ORs between 35.1 and 7.9 and P values between < 0.001 and 0.039).

CONCLUSIONS

This study again confirmed the prognostic impact of the criteria used in the Shimada system and revealed that some other morphologic features, such as prominent nucleoli in undifferentiated and poorly differentiated neuroblasts, identify unfavorable tumor biology, partly independent from the patient's age at diagnosis. However, the prognostic impact of these features needs to be confirmed by analysis of a large series of neuroblastic tumors. Cancer 2002;94:1574–83. © 2002 American Cancer Society.

DOI 10.1002/cncr.10359

Peripheral neuroblastic tumors (NTs; neuroblastoma, ganglioneuroblastoma, and ganglioneuroma) that evolve from immature, sympathetic neuroblasts during embryonic, fetal, or postnatal development constitute one of the most common categories of solid tumors in infancy and early childhood. The extensive search for prognostically significant morphologic, genetic, and clinical characteristics is attributable to the marked heterogeneity in clinical behavior observed in NTs that is expressed by a variety of genotypic and phenotypic features. The distinct genetic, pathologic, and clinical entities are of relevance for therapeutic decisions. For a long time, uncertainties existed in assessing the prognostic significance of various morphologic features of these tumors. Some of the difficulties reflected the primitive or undifferentiated nature of these tumors and the capacity in some tumors for spontaneous regression or maturation. Shimada and colleagues were the first to take the unique morphologic attributes of these tumors into account by unconventionally introducing the patient's age at the time of diagnosis into the assessment of the prognostic impact of histologic features.1 In addition, those authors also recognized for the first time that high cellular turnover, as determined by increased mitotic activity and karyorrhectic debris, is a sign of aggressive tumor behavior and that the development of a Schwann cell stroma is a condition sine qua non in the maturation process of NTs. Later, it was shown that Schwann cells do not belong to the neoplastic population but are nonneoplastic cells that most likely are recruited by neoplastic neuronal cells.2 Moreover, Schwann cells appear to be involved in the differentiation of neuroblastoma cells by the expression of neurotrophic factors (reviewed by Ambros and Ambros3 and Ambros et al.4). The prognostic significance of the presence of calcification and the number of mitotic figures was demonstrated by Joshi and coworkers, who also proposed an age-linked categorization based on these features.5 A modified histologic grading system using the number of mitotic and karyorrhectic cells (MKI) rather than the mitotic rate was suggested in 1996.6

The prognostic impact of the morphologic features included in the Shimada classification system and the consideration of the patients' age at diagnosis were confirmed by a review of 227 NTs.7, 8 This work resulted in an International Neuroblastoma Pathology Classification system proposed by the International Neuroblastoma Pathology Committee (INPC) with the adoption of the original Shimada system with minor modifications.7, 8

Members of the INPC conducted a new pathology review on a smaller series of neuroblastoma (Schwannian stroma-poor) tumors from two clinically defined subgroups of patients, i.e., those with clinically favorable tumors and clinically unfavorable tumors (for the definitions of these two groups, see Materials and Methods). The objectives of this study were to determine the biologic significance of various morphologic features for the discrimination of tumors in these two distinct clinical groups. The correlation between morphologic features and molecular/genotypic properties of NTs by using the same series of patients will be reported in a separate article.

MATERIALS AND METHODS

Hematoxylin and eosin-stained sections of 47 neuroblastoma (Schwannian stroma-poor) tumors from two distinct subgroups, i.e., patients with clinically favorable tumors (23 patients) and patients with clinically unfavorable tumors (24 patients), were reviewed in this study. Sections from tumors in the ganglioneuroblastoma, intermixed (Schwannian stroma-rich), ganglioneuroma (Schwannian stroma-dominant), and ganglioneuroblastoma, nodular (composite of Schwannian stroma-rich/stroma-dominant and stroma-poor) categories were excluded from the study, because, in these subtypes of neuroblastic tumors, distinctive individual cytologic features were not used for prognostic subgrouping as they are currently for tumors in the ganglioneuroblastoma, nodular category.9 The age of patients at the time of diagnosis ranged from 1 month to 102 months (median, 13 months). The clinically favorable group included patients with localized tumors (Stage 2A/2B, 17 patients; Stage 3, 4 patients) or with Stage 4 tumors (2 patients) without recurrence and/or progression after treatment with surgery or biopsy alone. It was determined that these patients had negative residual disease (one patient); positive residual disease (microscopic, two patients; macroscopic, seven patients); positive residual disease, not otherwise specified (six patients); and status unknown (seven patients) immediately after the surgery or biopsy. The clinically unfavorable group included patients with distant metastases (Stage 4, 13 patients) at the time of diagnosis; localized tumors at the time of diagnosis (Stage 1, 2 patients; Stage 2B, 2 patients; Stage 3, 3 patients) who eventually developed distant metastases during their clinical course; localized tumors (Stage 2A, 1 patient; Stage 3, 2 patients) that caused a fatal outcome after local tumor progression; and a Stage 4s tumor (1 patient) that was staged based on the initial clinical assessment but eventually was reclassified as Stage 4. In the clinically favorable group, patient age at the time of diagnosis ranged between 1 month and 76 months (median, 10 months). In the clinically unfavorable group, patient age at the time of diagnosis ranged between 2 months and 102 months (median, 20 months). It was noted that most patients age < 1.5 years at the time of diagnosis belonged to the clinically favorable group (72% of patients in this age group), whereas the majority of patients (76%) with NTs diagnosed at age > 1.5 years were in the clinically unfavorable group.

All microscopic sections (1–28 slides per patient; median, 4 slides per patient) from these 47 neuroblastomas were collected from 10 different medical institutions in Austria, Belgium, Germany, and Italy and were reviewed by 5 reviewers (I.M.A., J.H., V.V.J., B.R., and H.S.) without knowledge of clinical information. All tumor tissues were obtained prior to chemotherapy and/or irradiation therapy. Fourteen different morphologic features were analyzed in this study (Tables 1 and 2). Tumors from eight patients with localized tumors were derived from the Localized Neuroblastoma European Study Group: three tumors from Austria and five tumors from Italy. The features that were analyzed, such as grade of neuroblastic differentiation (neuroblastoma subtype), synchronous cytodifferentiation (identification for a differentiating neuroblast), development of ganglioneuromatous component, MKI, mitotic rate, and calcification, were defined by the INPC and have been described previously7 (for mitotic rate and calcification, see also Joshi et al.5). Definitions of other features, such as neuropil, cellularity, nuclear size, nuclear shape, prominent nucleoli in undifferentiated or poorly differentiated neuroblasts, nuclear inclusions of cytoplasmic material, tumor giant cells without pleomorphism, and tumor giant cells with pleomorphism, are listed in Table 1.

Table 1. Definitions of Morphologic Features Analyzed
Morphologic featureDefinitions
  1. HPF: high-power field.

NeuropilNo: the presence of neuropil cannot be identified unequivocally; minimal: single small, scattered foci of neuropil; sparse: more foci of neuropil or filamentous background that, nonetheless, is inconspicuous; moderate: neuroblasts are em- bedded and surrounded by clearly visible neuropil; prominent: large areas of the tumor are occupied by neuropil
CellularityLow: < 300 cells per HPF; intermixed: 300–600 cells per HPF; high: > 600 cells per HPF; intermediate: cellularity changes from low to intermixed and high
Nuclear sizeVariable, related to cytodifferentiation: small neuroblasts with nuclei with densely packed chromatin and inconspicuous nucleoli and scant cytoplasm are possibly found together with cells of all stages of synchronous cytodifferentiation (for definition, see Shimada et al., 1999a7); uniform: almost no variation in nuclear size can be observed; pleomorphic: variation in nuclear size which is not related to cytodifferentiation (according to the nuclear shape, the chromatin pattern and the lack of adequate cytoplasmic enlargement)
Nuclear shapeRound to oval: although the nuclei can differ in size, they do not deviate from a round to oval shape; pleomorphic: loss of the round-to-oval nuclear shape that, instead, is irregular (triangular, rectangular, polygonal, spindly, and bizarre)
Prominent nucleoli in undifferentiated or poorly, differentiated neuroblastsConspicuous, large eosinophilic nucleoli in small, medium-sized, or even large neuroblasts without signs of differentiation or maturation, such as synchronous cytodifferentiation (cytoplasmic enlargement), increase of neuropil, and decrease of cellularity
Nuclear inclusions of cytoplasmic materialInclusion of eosinophilic cytoplasmic material visible within the nucleus
Tumor giant cells without pleomorphismMultinucleated giant cells with mostly eccentric, round-to-oval nuclei of varying size and abundant eosinophilic cytoplasm
Tumor giant cells with pleomorphismMononucleated or multinucleated giant cells with a pleomorphic, often bizarre nuclear shape and a densely packed chromatin
Table 2. Categories of Features Used for Statistical Evaluation
Morphologic featureCategories
Neuroblastoma subtype (differentiation)DifferentiatingPoorly differentiated or undifferentiated
NeuropilProminent/moderateSparse or no/minimal
Synchronous cytodifferentiationPresentAbsent
Ganglioneuromatous componentPresentAbsent
CellularityLow/intermixed/intermediateHigh
Mitotic karyorrhectic indexLowIntermediate or high
Mitotic rateLowHigh
CalcificationPresentAbsent
Nuclear sizeVariable, related to cytodifferentiationUniform or pleomorphic
Nuclear shapeRound to ovalFocally or diffuse pleomorphic
Prominent nucleoli in undifferentiated or poorly differentiated neuroblastsAbsentPresent
Nuclear inclusions of cytoplasmic materialAbsentPresent
Tumor giant cells without pleomorphismSparse/manyAbsent
Tumor giant cells with pleomorphismSparseAbsent

The tumors initially were reviewed by each examiner and were then reviewed again as a group. After the individual review, analyses of the inter-reviewer agreement for each morphologic feature were performed. After the group review, consensus diagnoses (based on five of five or four of five agreements by the reviewers) were established for each morphologic feature and were used for further statistical analyses (see below). The analyses included a test for possible correlation between the individual morphologic features and tests for their prognostic effects.

The prognostic significance of each morphologic feature for distinguishing two clinical groups was analyzed and was then adjusted for the patient's age at diagnosis, and vice versa, because the latter had a very strong prognostic impact in this series. Prognostic significance of the INPC classification also was tested in this series.

To measure inter-reviewer agreement, the Cohen κ coefficient was applied based on data from the individual review diagnoses. All other statistical evaluations were based on the data from the consensus diagnoses. For statistical evaluation, each morphologic feature was divided into two categories (see Table 2). All correlations, except for the inter-reviewer agreement, were measured by odds ratios (ORs). To obtain conservative estimates, 0.5 was added to each cell, which is usual in log-linear models. For simple 2 × 2 tables, P values were computed using the chi-square test with Yates correction for continuity. To account for the confounding influence of covariates (like patient age) on the correlation of characteristics with clinical outcome in the stratified analysis, the Mantel–Haenszel method was applied, and the Mantel–Haenszel chi-square test was used. Interactions were tested by using the chi-square test for homogeneity of the OR. All P values are two-sided. P values ≤ 0.05 were regarded as significant. In accordance with the exploratory nature of this investigation, no correction for multiple testing was applied. Confidence intervals given are for a 95% level of confidence (95%CI).

RESULTS

Individual Morphologic Features

Inter-reviewer agreement

The features grade of neuroblastic differentiation (neuroblastoma subtype), synchronous cytodifferentiation, development of ganglioneuromatous component, prominent nucleoli in undifferentiated and poorly differentiated neuroblasts, and MKI reached the highest inter-reviewer agreement (κ, 0.42–0.53). By contrast, nuclear size, nuclear inclusions of cytoplasmic material, pleomorphic tumor giant cells, and nuclear shape showed the lowest values (κ, 0.10–0.2610).

Consensus diagnoses

In the group review, only three morphologic features, i.e., calcification, development of a ganglioneuromatous component, and nuclear inclusions of cytoplasmic material, reached 100% consensus. The assessment of cellularity, MKI, grade of neuroblastic differentiation (neuroblastoma subtype), amount of neuropil, and prominent nucleoli in undifferentiated and poorly differentiated neuroblasts showed a range of consensus between 95% and 90%. For the features synchronous cytodifferentiation, tumor giant cells without pleomorphism, nuclear size, mitotic rate, pleomorphic tumor giant cells, and nuclear shape, consensus was between 87% and 69%.

Correlations between the individual morphologic features

The highest degree of correlation was found between synchronous cytodifferentiation and other features, such as amount of neuropil, nuclear size, and tumor giant cells without pleomorphism. A correlation also was found between the features tumor giant cells without pleomorphism and nuclear size. There also were inverse correlations between cellularity and other features, such as amount of neuropil, synchronous cytodifferentiation, and nuclear size.

MKI was correlated with mitotic rate and was correlated inversely with nuclear size. Mitotic rate was correlated inversely with synchronous cytodifferentiation and nuclear size (also see Table 3).

Table 3. Correlations between Individual Morphologic Features, Clinical Group, and Shimada System Shown by the Odds Ratios and P Valuesa
VariablebClinical groupAge at diagnosisShimada systemNB subtypeNeuropilCytodifferent.GN comp.CellularityMK11MK12MRCalcificationNuclear sizeNuclear shapeNucleoliNuclear inclusionsGanglionic giant cellsPleom. giant cells
  • NB. neuroblastoma; Cytodifferent.: cytodifferentiation; GN comp.: ganglioneuromatous component; MK11: low versus intermediate and high mitotic karyorrhectic index; MKI2: low and intermediate versus high MKI; MR: mitotic rate; Pleom.: pleomorphic

  • a

    P values ≤ 0.05 were considered significant.

  • b

    For each variable, values to the left of the dash are odds ratios, and values to the right of the dash are P values.;1>

Clinical group0.0060.0000.2360.0020.0040.4010.0000.0030.0060.0190.0040.0000.0270.0000.1990.0041.000
Age at diagnosis7.40.0000.8090.0250.0620.6440.0070.8550.1790.8940.0040.1920.8360.0040.7270.0100.838
Shimada system17.319.50.4470.0020.0000.1260.0000.0160.0000.0130.0000.0020.8840.0010.2120.0000.333
NB subtype3.30.62.50.0210.0020.0250.0380.0080.1120.0170.9330.0010.6730.2730.8810.0160.678
Neuropil10.45.411.619.30.0000.1310.0000.0110.0020.0030.0120.0000.9080.0000.3160.0000.215
Cyto-different.9.74.421.741.9192.20.0450.0000.0030.0010.0020.0080.0000.6760.0010.2480.0000.300
GN comp.3.21.010.613.49.816.30.1430.0110.2810.0610.9610.0480.7090.8660.7920.0450.249
Cellularity26.47.422.616.0444.3217.89.30.0010.0000.0020.0010.0000.9430.0000.1110.0000.333
MK1110.51.47.513.610.716.725.744.30.0030.0000.5930.0010.1050.0120.6880.0930.976
MK129.63.193.210.114.221.66.120.131.00.0030.0160.0040.7270.0000.1640.0020.149
MR7.41.49.523.737.923.014.842.437.337.40.2020.0021.0000.0170.4360.0580.838
Calcification9.99.021.71.47.49.51.514.51.87.13.40.0580.6820.0030.0420.0020.082
Nuclear size36.93.220.461.795.3297.017.5155.029.935.029.45.70.5310.0030.3510.0000.510
Nuclear shape9.01.71.60.90.61.01.11.66.21.21.52.22.50.1970.6490.7700.028
Nucleoli21.79.522.43.634.923.51.751.210.624.411.010.617.44.40.0560.0010.414
Nuclear inclusions9.02.68.72.08.17.91.112.73.413.16.119.76.17.817.70.2480.723
Ganglionic giant cells9.78.025.722.964.1178.216.348.74.317.86.616.9130.20.923.57.90.029
Pleom. giant cells1.30.92.82.23.33.05.12.70.74.21.76.32.60.12.43.37.3

Prognostic Analysis

The distribution of the individual morphologic features in the favorable and unfavorable clinical groups and in the two age groups are summarized in Table 4, and the statistical evaluation (ORs, 95%CIs, and P values) is shown in Table 5.

Table 4. Distribution of Morphologic Features in the Clinical and Age Subgroups
Feature/age groupFavorableUnfavorable
  1. HPF: high-power fields.

Neuroblastoma subtype (differentiation)  
 Differentiating  
  Age < 1.5 yrs40
  Age > 1.5 yrs22
 Poorly differentiated  
  Age < 1.5 yrs137
  Age > 1.5 yrs19
 Undifferentiated  
  Age < 1.5 yrs00
  Age > 1.5 yrs02
Neuropil  
 Prominent  
  Age < 1.5 yrs40
  Age > 1.5 yrs10
 Moderate  
  Age < 1.5 yrs104
  Age > 1.5 yrs22
 Sparse  
  Age < 1.5 yrs30
  Age > 1.5 yrs04
 No or minimal  
  Age < 1.5 yrs03
  Age > 1.5 yrs06
Synchronous cytodifferentiation  
 Yes  
  Age < 1.5 yrs102
  Age > 1.5 yrs32
 No  
  Age < 1.5 yrs33
  Age > 1.5 yrs111
Ganglioneuromatous component  
 Present  
  Age < 1.5 yrs21
  Age > 1.5 yrs20
 Absent  
  Age < 1.5 yrs166
  Age > 1.5 yrs213
Cellularity  
 Low (< 300 cells per HPF)  
  Age < 1.5 yrs10
  Age > 1.5 yrs00
 Intermediate/intermixed (300–600 cells per HPF)  
  Age < 1.5 yrs162
  Age > 1.5 yrs32
 High (> 600 per HPF)  
  Age < 1.5 yrs23
  Age > 1.5 yrs011
Mitotic karyorrhectic index  
 Low (> 2%)  
  Age < 1.5 yrs91
  Age > 1.5 yrs41
 Intermediate (2–4%)  
  Age < 1.5 yrs63
  Age > 1.5 yrs03
 High (> 4%)  
  Age < 1.5 yrs23
  Age > 1.5 yrs07
Mitotic rate  
 Low (< 10 per 10 HPF)  
  Age < 1.5 yrs92
  Age > 1.5 yrs41
 High (> 10 per 10 HPF)  
  Age < 1.5 yrs54
  Age > 1.5 yrs06
Calcification
 Yes
  Age < 1.5 yrs175
  Age > 1.5 yrs34
 No
  Age < 1.5 yrs13
  Age > 1.5 yrs19
Nuclear size
 Variable (relative to cytodifferentiation)
  Age < 1.5 yrs90
  Age > 1.5 yrs31
 Uniform
  Age < 1.5 yrs23
  Age > 1.5 yrs19
 Pleomorphic
  Age < 1.5 yrs02
  Age > 1.5 yrs01
Nuclear shape
 Round to oval
  Age < 1.5 yrs120
  Age > 1.5 yrs34
 Focally pleomorphic
  Age < 1.5 yrs13
  Age > 1.5 yrs12
 Diffuse pleomorphic
  Age < 1.5 yrs00
  Age > 1.5 yrs01
Prominent nucleoli in undifferentiated or poorly differentiated neuroblasts
 No
  Age < 1.5 yrs164
  Age > 1.5 yrs33
 Yes
  Age < 1.5 yrs03
  Age > 1.5 yrs19
Nuclear inclusions of cytoplasmic material
 No
  Age < 1.5 yrs186
  Age > 1.5 yrs411
 Yes
  Age < 1.5 yrs01
  Age > 1.5 yrs02
Tumor giant cells, multinucleated without pleomorphism
 Sparse
  Age > 1.5 yrs102
  Age > 1.5 yrs00
 Many
  Age < 1.5 yrs11
  Age > 1.5 yrs21
 No
  Age < 1.5 yrs33
  Age > 1.5 yrs111
Tumor giant cells with pleomorphism
 Sparse
  Age < 1.5 yrs42
  Age > 1.5 yrs23
 No
  Age < 1.5 yrs83
  Age > 1.5 yrs17
Table 5. Odds Ratios, 95% Confidence Intervals, and P Values for the International Neuroblastoma Pathology Classification System, Patient Age at Diagnosis, Morphologic Characteristics, and Combined Analyses
FeatureOR95%CIP value
  • OR. odds ratio; 95% CI. 95% confidence interval.

  • a

    ORs of each feature were adjusted for patient age.

  • b

    ORs of patient's age were adjusted for morphologic feature.

Prognostic significance of the Shimada system and patient age at diagnosis in the neuroblastoma category
 Shimada system17.33.7–81.8< 0.001
 Age at diagnosis (1.5)7.41.9–28.90.002
Prognostic significance of morphologic features per se
 Nuclear size36.94.7–287.5< 0.001
 Cellularity26.44.9–142.6< 0.001
 Prominent nucleoli in undifferentiated or poorly differentiated neuroblasts21.73.3–143.60.001
 Mitotic karyorrhectic index10.52.2–51.00.002
 Neuropil10.42.4–45.70.001
 Calcification9.92.1–47.70.002
 Synchronous cytodifferentiation9.72.2–43.40.002
 Tumor giant cells without pleomorphism9.72.2–43.40.002
 Mitotic rate7.41.5–35.20.006
 Nuclear shape9.01.5–54.00.008
 Nuclear inclusions of cytoplasmic material9.00.4–186.00.078
 Neuroblastoma subtype (differentiation)3.30.7–16.60.072
 Ganglioneuromatous component3.20.4–22.30.124
 Tumor giant cells with pleomorphism1.30.3–5.50.358
Prognostic impact of morphologic features in addition to the prognostic information provided by patient age at diagnosisa
 Nuclear size35.13.7–336.6< 0.001
 Cellularity15.02.3–97.0< 0.001
 Mitotic karyorrhectic index13.82.0–97.30.004
 Prominent nucleoli in undifferentiated or poorly differentiated neuroblasts13.81.8–107.30.007
 Mitotic rate6.71.1–39.30.026
 Neuropil6.11.1–33.70.027
 Synchronous cytodifferentiation8.01.4–46.10.031
 Nuclear shape7.90.9–69.30.039
 Tumor giant cells without pleomorphism6.21.1–36.00.070
 Neuroblastoma subtype (differentiation)7.60.9–63.70.081
 Calcification6.81.1–43.40.082
 Nuclear inclusions of cytoplasmic material4.00.4–42.40.347
 Ganglioneuromatous component2.90.4–23.00.357
 Tumor giant cells with pleomorphism1.50.3–8.50.927
Prognostic impact of patient age at diagnosis in addition to the prognostic information provided by the morphologic featuresb
 Neuroblastoma subtype (differentiation)15.92.4–107.90.002
 Ganglioneuromatous component8.21.6–41.50.006
 Nuclear inclusions of cytoplasmic material7.01.7–29.20.009
 Mitotic karyorrhectic index7.90.9–69.00.017
 Tumor giant cells with pleomorphism7.51.3–44.00.031
 Nuclear shape5.80.6–57.60.065
 Neuropil4.50.8–27.30.080
 Synchronous cytodifferentiation5.71.0–33.70.093
 Calcification4.50.9–22.30.128
 Tumor giant cells without pleomorphism4.70.7–31.80.183
 Mitotic rate3.40.4–27.40.189
 Nuclear size5.20.6–43.80.264
 Cellularity3.40.5–25.90.291
 Prominent nucleoli in undifferentiated or poorly differentiated neuroblasts2.80.5–15.60.537

Prognostic impact of individual morphologic features

The morphologic features that showed the strongest correlations with the clinical groups were nuclear size, cellularity, prominent nucleoli in undifferentiated or poorly differentiated neuroblasts, MKI, amount of neuropil, calcification, synchronous cytodifferentiation, tumor giant cells without pleomorphism, mitotic rate, and nuclear shape, with individual ORs ranging from 36.9 to 9.0 and P values ranging between < 0.001 and 0.008 (see also Table 5). The other features, such as nuclear inclusions of cytoplasmic material, grade of neuroblastic differentiation (neuroblastoma subtype), development of ganglioneuromatous component, and tumor giant cells with pleomorphism, did not reach statistical significance. Nuclear inclusions of cytoplasmic material were encountered only rarely, but all three tumors with this feature were in the unfavorable category.

Prognostic impact of morphologic features adjusted for patient age at diagnosis and vice versa

Although patient age at diagnosis was a strong prognostic indicator by itself (OR, 7.4; 95%CI, 1.9–28.9; P = 0.002), eight morphologic characteristics retained significance after adjustment for age (see Table 5): Nuclear size, cellularity, MKI, and prominent nucleoli in undifferentiated and poorly differentiated neuroblasts showed the strongest prognostic impact in addition to the impact of patient age along with the features mitotic rate, neuropil, synchronous cytodifferentiation, and nuclear shape.

It also was noted that patient age at diagnosis provided additional significant power to the prognostic value of the following morphologic features: grade of neuroblastic differentiation (neuroblastoma subtype), development of ganglioneuromatous component, nuclear inclusions of cytoplasmic material, MKI < 2%, and tumor giant cells with pleomorphism. This was not the case for other nine histologic features (see Table 5). In contrast, the presence of prominent nucleoli in undifferentiated and poorly differentiated neuroblasts, high cellularity, and a uniform or pleomorphic nuclear size were correlated significantly with the clinically unfavorable group irrespective of the patient's age at diagnosis.

Prognostic significance of the INPC classification (Shimada) system

The prognostic significance of the INPC classification system had an OR of 17.3 (P < 0.001). Twenty tumors were assigned to the favorable histology group, and 19 tumors were assigned to the unfavorable histology group according to the INPC classification system. Eight tumors were not classified due to a lack of consensus diagnoses for one or more morphologic features. Sixteen of 20 tumors (80%) that were classified into the favorable histology group were in the clinically favorable group, whereas 16 of 19 tumors (84.2%) that were classified into the unfavorable histology group were in the clinically unfavorable group. An interesting aspect in such tumors diagnosed in patients age < 1.5 years was the source of more discordance (n = 6 patients) between histology groups and clinical groups compared with patients in the older age group (n = 1 patient). The INPC classification system added significant prognostic information in addition to the patient's age at diagnosis (OR, 9.4; 95%CI, 1.5–58.8; P = 0.015), with a higher OR (45.0) in the older age group compared with the younger age group (OR, 4.5).

DISCUSSION

Members of the INPC have continued to evaluate the classification scheme that was initiated in 1994 and resulted in its formal proposal in 1999.7, 8 The objective of the current histologic review was to test the reproducibility and prognostic impact of several standard morphologic characteristics together with other features for which the prognostic importance had not yet been analyzed in detail in tumors of the neuroblastoma (Schwannian stroma-poor) category.

Inter-Reviewer Agreement and Consensus Diagnoses

The highest inter-reviewer agreements were achieved for those features that have been in use for a long time in the pathologic diagnosis of neuroblastomas and that were included in the INPC classification system as well, such as the degree of neuroblastic differentiation, development of ganglioneuromatous component, synchronous cytodifferentiation, and MKI, which probably reflects the existence of clear definitions with good reproducibility. In addition to these features, it was found that the feature prominent nucleoli in undifferentiated and poorly differentiated neuroblasts had a high concordance rate among the reviewers. In contrast, features like nuclear shape, pleomorphic tumor giant cells, nuclear inclusions of cytoplasmic material, and nuclear size had lower concordance and consensus rates, reflecting the need for strengthened criteria and unequivocal definitions.

Prognostic Impact of Morphologic Features in Relation to Patient Age at Diagnosis

The individual morphologic features had varying degrees of prognostic significance. It also was noted that the feature patient age at diagnosis, by itself, had a strong prognostic impact in Schwannian stroma-poor neuroblastomas. The prognostic impact of the grade of neuroblastic differentiation (neuroblastoma subtype) became statistically significant only when it was considered along with patient age. Furthermore, the age at diagnosis provided prognostic information in addition to the features development of ganglioneuromatous component and MKI, all of which are used in the INPC classification system. This clearly reflects that some morphologic features in stroma-poor neuroblastomas have a different prognostic impact in different patient age groups, and it supports the concept of an age-linked prognostic classification. For example, as pointed out in our previous studies,7, 8 the maturation processes of neuroblastic cells may require a certain time (latent period) in vivo; primitive neoplastic cells with the potential for neuroblastic differentiation may not yet show recognizable features of morphologic differentiation in tumors that are excised or biopsied in the neonatal and early infancy period. However, regardless of the patient's age at diagnosis, the presence of prominent nucleoli in undifferentiated and poorly differentiated neuroblasts, high cellularity, and a uniform or pleomorphic nuclear size were correlated significantly with the clinically unfavorable group.

Morphologic Characteristics of Tumors in the Clinically Favorable Group

Clinically favorable neuroblastomas in patients age < 1.5 years usually (13 of 17 tumors; 76.5%) were examples of the poorly differentiated subtype with identifiable, often moderate amounts of neuropil and with an intermediate or intermixed cellularity. The remaining tumors (4 of 17 tumors; 23.5%) were classified into the differentiating subtype. It should be noted that clinically favorable neuroblastomas in patients in this age group were not of the undifferentiated subtype with a complete lack of appreciable neuropil. All tumors showed a variation in nuclear size related to cytodifferentiation, and the differentiating neuroblasts, although they did not usually exceed 5% of the total cell population, were found in the majority of tumors. These tumors rarely (2 of 17 tumors; 11.8%) had a high MKI. In contrast, a high mitotic rate was seen in 5 of 15 clinically favorable tumors (35.7%) in this age group. The nuclear shape was almost always round to oval without pleomorphic features. No prominent nucleoli were seen in the neuroblastic cells in any of these tumors. Tumor giant cells without pleomorphism were encountered frequently. Calcifications were identified in all but one tumor.

Clinically favorable neuroblastomas were infrequent (four tumors) in patients age > 1.5 years in this series of tumors, which did not include Schwann cell stroma-rich tumors. Two of three tumors were classified into the differentiating subtype due to the grade of neuroblastic differentiation. Most importantly, all tumors showed a low MKI, a low mitotic rate, an intermediate or intermixed cellularity, and moderate or prominent amounts of neuropil. It was found that one differentiating neuroblastoma had prominent nucleoli in the less differentiated neuroblastic population within the same tumor tissue (see also below).

Morphologic Characteristics of Tumors in the Clinically Unfavorable Group

In this group of tumors, a combination of a high MKI, high cellularity, and presence of prominent nucleoli in undifferentiated and poorly differentiated neuroblasts exclusively identified tumors in the clinically unfavorable group regardless of the patient's age. In contrast, the two clinically favorable neuroblastomas with a high MKI showed moderate amounts of neuropil, an absence of high cellularity, and a lack of prominent nucleoli and were diagnosed in patients age < 1.5 years. In patients age > 1.5 years, a high MKI and a high mitotic rate identified clinically unfavorable tumors. High cellularity was always an adverse prognostic sign in tumors from children age > 1 year (11 tumors). Without exception, the absence of neuropil or the presence of only minimal amounts of neuropil (9 tumors) was found only in clinically unfavorable tumors irrespective of the patient's age at diagnosis. There were two undifferentiated tumors: Both were diagnosed in patients age > 1.5 years. These two tumors, by definition, did not have any appreciable amount of neuropil.

It was noted that 12 of 13 tumors (92.3%) with prominent nucleoli in undifferentiated and poorly differentiated neuroblasts belonged to the clinically unfavorable group. Patients with these unfavorable tumors had either a high MKI (10 patients) or an intermediate MKI (2 patients). Only one exception was a clinically favorable tumor of the differentiating subtype with prominent nuclei in the less differentiated neuroblastic population (see above) that had a low MKI. The presence of prominent nucleoli in undifferentiated and poorly differentiated neuroblasts and increased mitotic and karyorrhectic activities apparently were linked to one another (see Table 3). It has been reported that increased MKI, one of the indicators of an unfavorable neuroblastoma in the INPC classification system, may be one of the morphologic manifestations of MYCN amplification in neuroblastic tumors.10 It is noteworthy that it was shown that the myc RNA (N-myc as well as c-myc) localizes to the nucleolus,11 a fact that could explain the pronounced enlargement of nucleoli in undifferentiated or poorly differentiated cells by the accumulation of N-myc messenger RNA.

Nuclear inclusions of cytoplasmic material, most likely a sign of rapid cell proliferation, were observed in only three tumors from the clinically unfavorable group combined with a high MKI and prominent nucleoli. The nuclear size in most of the unfavorable tumors was either uniform or pleomorphic, with one exception: a neuroblastoma of the differentiating subtype that was diagnosed in a child age > 6 years. In contrast to the tumors from the favorable clinical group, the majority of the unfavorable tumors had no nonpleomorphic tumor giant cells.

Lack of calcifications usually was associated with aggressive tumor behavior. However, at the same time, 43% of the unfavorable tumors showed evidence of calcification.

The International Neuroblastoma Pathology Classification (Shimada) System and Other Morphologic Features in the Prognostic Evaluation of Tumors in the Neuroblastoma (Schwannian stroma-poor) Category

The current review has demonstrated the prognostic significance of the INPC classification (Shimada) system in evaluating Schwannian stroma-poor neuroblastomas, in that it takes into account the degree of neuroblastic differentiation, the number of mitotic and karyorrhectic cells, and patient age at the time of diagnosis. Calcification and mitotic rate, the indicators used by Joshi and coworkers, also had individual prognostic significance, although they did not add any prognostic information to the Shimada system. Beyond this, some other morphologic features, such as prominent nucleoli in undifferentiated and poorly differentiated neuroblasts, cellularity, and nuclear size, were identified to be of significant clinical relevance irrespective of the patient's age at diagnosis. Whether these features or other morphologic features investigated in this study can add prognostic information significantly to the Shimada system has to be clarified by evaluating a larger series of patients with NT together with checking their reproducibility.

Acknowledgements

The authors thank D. Modritz for secretarial help and assistance with data management and G. Pascher and M. Zavadil for their critical reading of the article.

Ancillary