3

Over 125 years have passed since the historic report of Bennett and Godlee1 in 1888 describing their initial surgical approach to a patient with a primary brain tumor; yet, the surgical aspects of malignant glioma remain an area of discussion and controversy. The role of surgical intervention continually evolves as new techniques become available and continued study outlines the effectiveness and contributions of surgical procedures for survival and quality of life for patients with malignant glioma. Controversy has arisen partly because of the lack of any consistent effective therapeutic intervention. Numerous factors have been proposed as contributors to the overall outcome in this patient population.

Primary malignant brain tumors are among the most difficult human malignancies to treat. In contrast to small metastatic lesions to the brain, small malignant primary tumors often progress rapidly despite multi-modality therapy. A recent review estimated 18,400 cases of central nervous system cancer will occur in 2004; 12,690 deaths will be attributed to the tumors.² The majority of these primary central nervous system cancers are malignant gliomas. Despite recent evidence demonstrating improved survival over the past 30 years, the possibility of long-term survival remains remote and treatment decisions are focused on neurocognitive and quality-of-life outcomes.² In many instances, the main bulk of the tumor is managed differently than the remaining rim of infiltrating malignant cells that are invariably present. A significant challenge remains in developing therapies targeted at this infiltrative component. This residual area, usually outside the zone of enhancement, is the site of local recurrence and is more likely to cause problems with long-term control.

In this chapter, we review some clinical aspects of malignant glioma to demonstrate that the principles of evidence-based medicine can be successfully applied to the study of brain tumors. Emphasis is placed on examining the availability of reasonably high-quality medical evidence to guide therapeutic decisions. Areas of focus include patient assessment (the reliability of histologic diagnosis of malignant glioma), establishing the diagnosis (the accuracy of stereotactic biopsy in malignant glioma), determining of the prognosis (the natural history of malignant glioma), and evaluating of the intervention (the role of cytoreductive surgery).

In our research, we have used the National Library of Medicine’s MEDLINE database. Additional resources, however, for evidence-based neuro-oncology studies include the Cochrane Reviews (available through their Web site at www.cochrane.org) and Neurosurgical Focus, which has published evidence-based articles and guideline development projects of the American Association of Neurological Surgeons and the Congress of Neurological Surgeons (AANS/CNS) Joint Section on Tumors. Neurosurgical Focus is available through their Web site at www.aans.org/education/journal/neurosurgical.

Patient Assessment

Reliability of Histologic Diagnosis of Malignant Glioma

The diagnosis of a malignant glioma continues to be made on histopathologic analysis; tissue obtained either at the time of biopsy or cytoreductive surgery is evaluated (Fig. 3–1). To evaluate the diagnostic accuracy in patients with malignant glioma, we initiated a MEDLINE review of the medical literature using the keyword glioma and cross-referenced with biopsy, accuracy, diagnosis, variability, and pathology. Initially, we reviewed the titles and abstracts to remove duplicates and eliminate erroneous entries; we then scanned bibliographies of selected articles for additional relevant references. We identified numerous studies on noninvasive imaging techniques for refining the diagnosis, but our focus in this chapter is on those studies addressing the actual analysis of the tissue. We selected several articles that specifically cover the techniques of diagnosing malignant glioma and outline them below. The studies include issues of observer variability, the contribution of cytologic techniques, and the role of image analysis (Table 3-1).

Mittler et al³ reported an evaluation of observer reliability in the histologic grading of glial tumors. In this study, four neuropathologists under-went a blinded review of 30 brain biopsy specimens. Both intraobserver and interobserver agreements were analyzed. Intraobserver agreement was 75% for glioblastoma (GBM) and 51% for anaplastic astrocytoma (AA) with a mean intraobserver agreement of 63.9% (kappa = 0.50). Interobserver agreement was 62% for glioblastoma (kappa = 0.39) and 36% for AA (kappa = 0 .06). The authors pointed out the considerable discrepancy observed, not only between different pathologists but also between interpretations of the same pathologists at different times.

Figure 3–1 Histologic section of glioblastoma.

A subsequent blinded study by Prayson et al4 compared the grading consistency of five neuropathologists with five surgical pathologists in retrospectively evaluating 30 biopsy specimens of normal brain and varying grades of glial tumors. Agreement among the neuropathologists was complete (five of five) in 12 of 30 cases (40%), and four of the five agreed in 26 of 30 cases (87%). In contrast, there were only six cases (20%) in which the surgical pathologists all agreed, 13 cases (43%) where four of the five surgical pathologists agreed. At least two disagreed with the rest of the group in 50% of the cases. The kappa value for interobserver variability among the neuropathologists was reported as higher than that for the surgical pathologists (kappa = 0.63 versus 0.36). From the data provided in the article, the kappa value for agreement on the diagnosis of glioblastoma among the neuropathologists was calculated to be 0.81 versus 0.63 for the surgical pathologists and 0.88 versus 0.55 for AA. The authors conclude that experience in grading glial tumors is an important factor in reliably establishing the diagnosis, and that it can be accomplished in a high percentage of cases. Additional observation suggested agreement on the diagnosis of the malignant tumors (GBM and AA) is less variable than in the entire spectrum of glial neoplasms included in this study.

Roessler et al⁵ retrospectively reviewed 4172 intraoperative cytologic smears from intracranial procedures (3541 open procedures and 631 stereotactic biopsies with a wide variety of diagnoses) and compared preliminary versus final pathological diagnosis. Complete correlation was obtained in 90%; the correlation in the diagnosis of glioblastoma was reported as 95.7% with a calculated sensitivity of 0.97 and specificity of 0.99. Significantly less diagnostic accuracy was achieved with lower-grade glial tumors (oligodendroglioma = 81% and ependymoma = 78%). The authors conclude that cytologic smears have a high diagnostic accuracy when compared retrospectively with the final histopathologic diagnosis in the majority of cases and are particularly accurate in high-grade glial tumors. In this study, the glioblastoma group appeared to include all malignant glial tumors (combining both AA and GBM).

Abbreviations: AA, anaplastic astrocytoma; AST, astrocytoma; GBM, glioblastoma; MIB–1 LI, MIB-1 labelling index

Coons et al⁶ reported 244 cases of glioma that were reviewed by four neuropathologists in four separate sessions to evaluate interobserver variability. Significant improvement in diagnostic concordance with each session was observed (p = 0.02). For the first session, the concordance rates were as follows: all four reviewers, 52%; any three reviewers, 60%; two reviewers, 70%. By the fourth session, this rate improved to 69%, 75%, and 80%, respectively. Kappa values were not reported and could not be calculated.

Sharma et al⁷ evaluated 43 cases of astrocytic tumors and mixed gliomas. They compared the reproducibility of the Kernohan grading system with a computer-aided malignancy classifier. High inter- and intraobserver variability were observed using the Kernohan grading system.⁸ However, morphometric evaluation of four histological variables in the classifier revealed a statistically significant difference between the clusters of the measured quantitative values. Repeated grading of histological features using a computer assisted classifier resulted in complete elimination of interobserver variability. The authors concluded that reproducible morphometric techniques might be of use in the grading of gliomas. This was an encouraging preliminary study on a small number of cases, but the limited amount of data presented precluded additional analysis.

Gaudin et al⁹ analyzed the diagnostic accuracy of a combination of cytopathologic and histopathologic techniques in a series of 74 patients undergoing stereotactic biopsy. Included in this study was a comparison of the crush-prep cytology technique with the frozen-section diagnosis. For those cases diagnosed as GBM, the sensitivity and specificity of the cytological technique was 0.54 and 1.0, respectively. For AA cases, the sensitivity was 0.81 and the specificity was 1.0. In a multivariate survival analysis, three factors were correlated with decreased survival: age >55 years [p < 0.001, hazard ratio = 3.58; confidence interval (CI) = 1.95, 6.57], nuclear atypia (p = 0.004, hazard ratio = 4.30; CI = 1.61, 11.46), and necrosis (p = 0.016, hazard ratio 2.14; CI = 1.15, 3.97). These authors propose that a high correlation between the cytologic crush-stain specimens and the final diagnosis could be demonstrated, and that presence of nuclear atypia and necrosis on the initial specimen could be strongly correlated with survival.

Marie et al10 described a series of 50 astrocytic gliomas and determined the interobserver variability of MIB-1 labelling index (MIB-1 LI) calculated by six pathologists using cut-off values of 2.5, 5.0, 8.0, 11.0, and 15.0%. The interobserver variability was high; the best results were with a cut-off value of 5.0% yielding a pairwise kappa value ranging from 0.52 to 0.80. The authors conclude the high level of interobserver variability suggested MIB-1 LI prognostic cutoff values might not be clinically useful for predicting outcome in individual patients with primary brain tumors. Further prospective studies are needed for investigating the prognostic usefulness of MIB-1 LI ranges that optimize interobserver agreement.

Revesz et al¹¹ described 273 patients with gliomas and compared pathological features with survival. Data were available on 209 of 273 patients (76%). Grading, using both the Kernohan criteria and Daumas–Duport criteria,¹¹ was compared with survival and specific pathological features examined for their contribution.¹¹ Findings suggested both the Kernohan and Daumas–Duport classification systems significantly correlated with survival (Cox analysis, p < 0.0001), but the Kernohan did not differentiate between grades III and IV, while the Daumas–Duport did. Mitosis (p < 0.0001, x² = 17.9), endothelial proliferation (p < 0.0001, x² = 39.4), and necrosis (p = 0.0007, x² = 11.5) were all significantly correlated with survival.

Martin et al¹² evaluated a system of image analysis to examine 272 glioma specimens including 70 GBMs, 91 astrocytomas (ASTs), 56 pilocytic astrocytomas, and 55 oligodendrogliomas. A series of morphological properties of the nuclei and mitotic parameters were assessed, and an automated scale was created. A high degree of correlation was noted using standard grading methodology with statistically significant relationships described between numerical nuclear density, deformation of nuclei, and mitotic activity (Student’s t test, Wilcoxon’s test α = 0.05). Despite promising initial results, no data relating these findings to survival are given; hence, the sensitivity and specificity of the automated testing cannot be determined.

The study by Decaestecker et al¹³ proposed to incorporate computer-assisted microscopy and artificial intelligence methodology to improve the accuracy of prognostic values associated with astrocytic tumors. A series of 250 patients with glial tumors (164 GBMs, 47 AAs, and 39 astrocytomas) were included. This study demonstrated that computer-aided methodology could potentially improve conventional methods for determining prognosis based on morphologically-based classifications and suggested methodology for further refining traditional classifications into new clinically useful subgroups. The authors detected a small but significant correlation between survival and a new variable under study: the percentage of hypertriploid cell nuclei. The two factors were considered significantly correlated (r = 0.35, p < 0.0001). From the data given, the sensitivity and specificity of the test were unable to be determined.

SUMMARY

Establishing the Diagnosis

The Accuracy of Stereotactic Biopsy in Malignant Glioma

The diagnosis of a malignant glioma is often made based on the limited tissue available in a stereotactic biopsy (Fig. 3–2). To evaluate the diagnostic accuracy of stereotactic biopsy in patients with malignant glioma, we did a MEDLINE review of the medical literature using the keyword glioma and cross-referenced with biopsy, accuracy, diagnosis, variability, and pathology. As before, titles and abstracts were reviewed to remove duplicates and eliminate erroneous entries; the bibliographies of selected articles were also reviewed for additional relevant references. We selected a series of reports addressing the diagnostic accuracy of stereotactic biopsy for diagnosing malignant glioma; they are outlined below. The articles we identified focused on cases in which the intraoperative diagnosis (cytological evaluation or frozen section) could be compared with the final pathological diagnosis (open craniotomy or autopsy).

Figure 3–2 Patient prepared for image-guided stereotactic biopsy.

A number of studies have addressed the diagnostic accuracy of stereotactic biopsy in the initial management of patients suspected of harboring a malignant glioma. In the studies reviewed below, accuracy has been reported as ranging from 62 to 95%, but significant variability exists in the number of patients included and the methodology involved in these studies. The issue is one that lends itself to a well-designed study. For example, if the study includes all patients undergoing a diagnostic biopsy, who were later confirmed by craniotomy or autopsy to either have or not have a malignant glioma, the sensitivity, specificity, and positive and negative predictive value can all be calculated (Table 3-2).

Kleihues et al¹⁴ reported an early study addressing the validity of morphological diagnosis of stereotactic brain tumor biopsies. This study drew from a series of 600 patients undergoing stereotactic biopsy for presumed brain tumors. Confirmatory pathology was available by open craniotomy or autopsy in 87 cases. Cytological (smear preparations) and histological examinations of paraffin-embedded samples were compared. The biopsy diagnosis was accurate in 67 of 71 cases in which a tumor was diagnosed with a sensitivity of 0.94 and a specificity of 0.64. The authors believed inaccurate diagnoses resulted from sampling errors in nonhomogeneous tumors; that is biopsies taken from sites not representative for the entire neoplasm (tumor necrosis, infiltration zone).

Chandrasoma et al¹⁵ described a retrospective review of the diagnostic accuracy of stereotactic biopsy in 30 patients who subsequently under-went open craniotomy and compared the intraoperative diagnosis with the final diagnosis. The correlation between the stereotactic and final diagnosis was exact in 19 of 30 cases and indicated a correct clinical treatment in an additional nine cases (overall appropriate for therapy in 28 of 30 cases). There were no false-positive results for tumor and two false-negative results: One patient with glioblastoma (GBM) whose stereotactic biopsy demonstrated necrotic tissue and one patient diagnosed with granulomatous inflammation on stereotactic biopsy who was later diagnosed with a pineal germinoma. The sensitivity for the presence of tumor was 0.92 and the specificity was 1.0. The authors conclude that stereotactic technique can provide biopsy material representing the lesion with sufficient accuracy for clinical management in the majority of cases.

Feiden et al¹⁶ reported their series of 47 patients undergoing stereotactic biopsy for intracranial mass lesions with a subsequent diagnosis made at either craniotomy (38 patients) or autopsy (9 patients). The final diagnoses included low- and high-grade glial tumors as well as lymphoma, metastasis, and inflammation (aspergillosis, toxoplasmosis). In 42 cases (89%), histological results from the stereotactic biopsy and resection/autopsy were identical. The study indicated specificity for the presence of tumor of 0.87 and a specificity of 1.0. It was not possible from the data presented to calculate these values for the glial tumors alone. The authors conclude that stereotactic biopsy results in a high level of diagnostic accuracy.

Colbassani et al17 reviewed 97 patients undergoing stereotactic biopsy for diagnosis and compared frozen section with final diagnosis. In 78 cases, the frozen-section diagnosis matched the permanent diagnosis. There were five false-negative results and one false-positive result. The sensitivity was 0.94 and the specificity was 0.93. The authors conclude that intraoperative frozen sectioning is a useful and highly accurate technique to ensure appropriate tissue is obtained during a stereotactic procedure and that it results in an accurate diagnosis in a high number of cases.

Brainard et al¹⁸ conducted a retrospective review of 188 stereotactic biopsies for intracranial pathology. This study compared the results of 177 of the cases with intraoperative frozen section to the final paraffin section diagnosis. In the overall group, there was one false-positive result and 35 false-negative results. The sensitivity was 0.73 and the specificity was 0.97. The authors proposed multiple biopsies be taken until diagnostic tissue is obtained and confirmed by frozen section.

Roessler et al⁵ performed a retrospective analysis of cytological smears from 3541 craniotomies and 631 stereotactic biopsies. Correlation of the smear with the final diagnosis was achieved in a mean of 89.8% (range, 83 to 93.7% per year). Diagnostic accuracy increased to 95% on average (range, 91.5 to 96.7% per year) when cases of partial correlation mainly due to grading deviations were included. The accuracy of intraoperative diagnoses for glioblastoma was 95.7%. A significant reduction in diagnostic accuracy was observed in cases of oligodendroglioma (80.9%) and ependymoma (77.7%). For GBM the sensitivity was 0.97 and the specificity was 0.99. The authors note intraoperative smears are inexpensive and have high diagnostic accuracy.

Gaudin et al⁹ conducted a blinded review of 27 GBM and 21 AA cases comparing diagnosis based on smear cytology with final diagnosis. For glioblastoma, the cytological diagnosis had a sensitivity of 0.54 and a specificity of 1.0. For AA, the sensitivity of the cytological smear was 0.81 and the specificity was 1.0. From this small subgroup of patients, the cytological smear was effective in excluding the diagnosis of a malignant glioma (AA or GBM), but was less sensitive for the diagnosis of GBM than for AA.

Abbreviations: AA, anaplastic astrocytoma; AST, astrocytoma; GBM, glioblastoma

Silverman et al19 reported 31 cases of stereotactic biopsy followed by craniotomy in which the cytological diagnosis could be compared with the final histological diagnosis. This series included 13 cases of malignant glioma (three AAs and 10 GBMs). Overall, the sensitivity for detecting a neoplasm was reported as 0.97 with a specificity of 1.00. Considering malignant glioma alone (AA as equivalent to GBM), the sensitivity was 0.85 and specificity was 1.0. Considering AA as distinct from GBM, the sensitivity was 0.70 and the specificity was 0.90.

In a study of 650 cases reported by Bleggi–Torres et al,²⁰ the authors compared the cytological versus paraffin-fixed diagnosis prepared on the same specimen at the time of sampling. The accuracy of the smear was 97.3% with a sensitivity of 0.98 and a specificity of 0.95. The authors report a positive predictive value of 0.99 and a negative predictive value of 0.90. There was one false-negative result and two false-positive results. The authors concluded the diagnosis based on cytological preparation has a high correlation with paraffin-embedded specimens.

Revesz et al¹¹ reviewed 419 stereotactic biopsy results, including 273 patients with gliomas, comparing the intraoperative diagnosis with the final diagnosis. For the diagnosis of a glial tumor by stereotactic biopsy using both cytology smear and a paraffin section, this study described a sensitivity of 0.92 and a specificity of 0.99. There were two false-positives and 23 false-negative cases in the overall group.

Shah et al²¹ studied 183 patients with central nervous system tumors. The investigators compared intraoperative cytology with frozen section. The diagnostic accuracy for the cytology smear was 89.7 and 90.4% for the frozen section. Overall, for the diagnosis of high-grade glioma described in this study, they found a sensitivity of 0.88 and a specificity of 0.99.

Firlik et al22 retrospectively reviewed 595 cases of stereotactic brain biopsy. The intraoperative cytological diagnosis correlated with the final diagnosis in 90% of cases. A diagnosis could be made in 91% of cases with sensitivity of 96% and a specificity of 75%.

Jackson et al²³ retrospectively reviewed a consecutive series of 81 patients whose imaging studies suggested a glioma and who then underwent stereotactic biopsy followed by craniotomy. Outside pathologists first examined the specimens; three neuropathologists at the study institution then reinterpreted their diagnoses. Comparison of the diagnoses between the stereotactic and open specimen differed in 49% when using the diagnosis from the outside pathologist. This discrepancy was reduced to 38% when each of three neuropathologists at the study institution reviewed the biopsy slides preoperatively. No further statements on interobserver or intraobserver variability are possible from the data presented. In cases with a stereotactic biopsy diagnosis of GBM, the sensitivity was 0.81 and the specificity was 1.0; for AA, the values were 0.67 and 0.64, respectively. The authors expressed concern that stereotactic biopsy can frequently lead to an inaccurate diagnosis and should be interpreted by pathologists experienced in the grading of glial tumors. In this study, it appeared the stereotactic biopsy diagnosis of glioblastoma was associated with a higher diagnostic accuracy than AA.

SUMMARY

Determination of Prognosis

Natural History Studies

MALIGNANT GLIOMA

The prognosis of a patient with a newly diagnosed malignant glioma remains grim (Fig. 3–3). Because most patients receive some type of therapy once the diagnosis is made, information on the natural history is difficult to obtain. Despite advances in neuroimaging, surgical technique, genetic analysis, radiation therapy, and chemotherapy, improvement in survival has been minimal. According to historical controls reviewed by the American Cancer Society, the 5-year survival rate for CNS cancer (the majority being glial in origin) from 1974 to 1976 was 22%, from 1983 to 1985 was 27%, and from 1997 to 1999 was 33%. The improvement in survival was statistically significant when comparing the 1974 to 1976 data to the 1997 to 1999 data (p < 0.05).² Despite overwhelming odds, a limited number of patients appeared to achieve a long-term survival. In evaluating the prognosis of malignant glioma, we examined the available literature on the natural history and the potential of a cure in this patient population. In our MEDLINE review, we used the keywords glioma, natural history, cure, and prognosis. Initially, we evaluated titles and abstracts to remove duplicates and eliminate erroneous entries. We reviewed bibliographies of selected articles to accumulate additional articles for consideration. Although limited in scope, several articles contained information addressing the natural history of malignant glioma and glioblastoma, and while anecdotal in nature, studies were identified that described experience with long-term survivors; we summarize them below.

< div class='tao-gold-member'>

Only gold members can continue reading. Log In or Register to continue

Related posts:

Evidence-Based Neurosurgery in Practice: Blending Art and Science Evidence-Based Medicine: A Conceptual Framework Cerebrovascular Surgery Peripheral Nerve Surgery Radiosurgery for Brain Metastases Management of Neurotrauma

Stay updated, free articles. Join our Telegram channel

Join