Differential Diagnosis

An enlarging lesion at the site of a previously treated tumor likely represents renewed growth of an incompletely eradicated initial tumor rather than the development of a new pathologic entity. Exceptions are infrequent but those in the following list need to be considered:

These alternative diagnoses must be excluded before prognosis is addressed and therapy is chosen. Neurodiagnostic imaging usually permits accurate prediction of the diagnosis. Generally, recurrent tumors have imaging features similar to those of the original lesion. A recurrent malignant glioma will likely have central low intensity, rim enhancement, and hypointense surround on enhanced T1-weighted magnetic resonance imaging (MRI). In some cases, however, attention to subtle differences may be required: a more spherical, sharply demarcated shape may suggest abscess rather than recurrent malignant glioma; and a diffuse, irregularly marginated pattern of surrounding edema may indicate radiation necrosis rather than recurrent tumor.

Two scenarios, malignant progression and radiation effects, often pose particular diagnostic difficulty. In each case, alternative diagnoses are often impossible to distinguish using current imaging modalities alone. Thus, biopsy for histologic evaluation and confirmation may be necessary.

Malignant Progression

The first scenario is the renewed growth of a low grade tumor. When low grade gliomas regrow after therapy, approximately half remain nonanaplastic, but the other 50% have progressed to a more malignant form.³ Molecular analyses have delineated genetic correlates of this progression.⁴ Enlarging low grade tumors will usually resemble the original tumor on imaging studies. When progression in grade has occurred, the new tumor may also resemble the old one, especially if the original tumor enhanced with contrast. Enhancement is highly predictive of likelihood of recurrence; low grade enhancing tumors are 6–8 times more likely to recur than nonenhancing ones.³ Most commonly, new malignant growth in a previously nonenhancing glioma enhances and thus is readily identified. In one study, only 30% (16/42) of low-grade tumors enhanced initially, but 92% (22/24) enhanced at recurrence.³ Occasionally, an enlarging malignant focus may not enhance. It might, however, be apparent as a region of hypermetabolism on a 2-deoxyglucose or 11-C methionine PET study, or have an increased rate of enhancement on a dynamic MRI scan, increased activity on a dual-isotope, single-photon emission computerized tomogram (SPECT), or increased choline signal on magnetic resonance spectroscopy (MRS).^5–7 The differential specificity of each of these new modalities is approximately 80% to 90%. Usually, however, histologic analysis after biopsy or resection is warranted to verify malignant transformation.⁸

Radiation Effects

The second scenario that causes diagnostic difficulty is renewed enlargement of a tumor mass following radiation. Often, CT and MR imaging inadequately distinguish recurrent tumor from radiation-induced enlargement. Usually only large, very malignant tumors grow sufficiently fast to show significant enlargement during, or within 3 months of completing, a course of radiation. When this does occur, the prognosis is particularly poor.⁹

Radiation can cause tumor enlargement in three ways:¹ through an early reaction, occurring during or shortly after irradiation, which is likely to be edema;² through an early delayed reaction arising a few weeks to a few months after radiation which involves edema and demyelination; and³ through a late delayed reaction that occurs 6 to 24 months after radiation and reflects radiation-induced necrosis.¹⁰ Regional teletherapy to a dose of 60 Gy is the current standard radiation treatment for most gliomas.¹¹ Although these doses have a low risk of inducing radiation necrosis, regional early and early delayed effects are relatively common. In most cases, tissue swelling represents edema and is transient. Acute symptoms from early and early delayed effects of radiation usually respond quickly to a short course of corticosteroids. The low density, T1 hypointense, T2 hyperintense regions of edema correspond to the area irradiated. Chronically, these volumes of brain will demonstrate parenchymal atrophy, enlargement of subarachnoid spaces, and ex vacuo ventricular dilatation. Dementia with apathy, inanition, and memory loss and decline in fine motor control are the clinical correlates. In the absence of new tumor growth, enhancement on CT and MRI beyond the initial resection margin is infrequent; when it does occur, it is patchy, irregularly marginated, and it can be distinguished from the more focal appearance of recurrent tumor.

In contrast, the late delayed effect of radiation-induced necrosis appears at about the time malignant tumors might be expected to recur.¹² It is thus more likely to be mistaken for recurrent tumor growth. The risk of radiation necrosis increases with the volume of tissue treated, the dose delivered, and the fraction size.¹³ Radiation necrosis following fractionated treatment to doses less than 70 Gy is rare, but it is much more common following brachytherapy or radiosurgery, which deliver high doses of radiation to relatively small volumes over a short time period.^12,14,15 A common protocol for brachytherapy is a 50- to 60-Gy boost (to 60 Gy of regional external beam radiotherapy) to a 0- to 5-cm tumor delivered over approximately 1 week. The radiosurgery equivalent is a 10- to 20-Gy boost to a 0- to 3-cm diameter tumor delivered in less than 1 hour.¹⁶ Necrosis is radiographically and pathologically evident in almost all cases and symptomatic in about half.

Whether it arises from higher doses of fractionated radiotherapy, brachytherapy, or radiosurgery, radiation necrosis is often difficult to distinguish from recurrent tumor. It forms a ring contrast-enhancing mass that resembles a malignant tumor. It has a CT hypodense, T1 hypointense, T2 hyperintense center; an enhancing annular region; and a hypodense, T1 hypointense, T2 hyperintense surround. The surround corresponds to edema that strikingly radiates along white matter tracts. The similarity of this appearance to that of recurrent tumors and the time course of its occurrence frequently necessitates additional measures to differentiate radiation-induced necrosis from recurrent tumor. A variety of functional neurodiagnostic imaging techniques attempt to distinguish between these two possibilities. These include PET scans, SPECT studies, cerebral blood volume mapping, and MRS. Regions of high activity are thought to distinguish recurrent tumor from relatively metabolically inactive and hypovascular radiation necrosis.⁶ Although specificity in differentiating tumor from radiation necrosis of up to 100% has been claimed, in many cases these studies are inconclusive and the diagnosis is revealed either by the clinical course or by analysis of a pathology specimen.

When an enlarging mass, which is either recurrent tumor, radiation necrosis, or both, becomes symptomatic, corticosteroid therapy is required.¹⁷ Up to half the patients receiving brachytherapy and radiosurgery for a malignant glioma develop symptoms that either prove refractory to corticosteroids or require debilitating long-term steroid use.^12,18,19 Surgery for resection of an enlarging, symptomatic mass is needed in 20% to 40% of cases following brachytherapy or radiosurgery of a malignant glioma. At reoperation for presumed radiation necrosis following focal radiation treatment of a malignant glioma, necrosis without tumor was found in 5% of cases, tumor alone in 29%, and a mixture of radiation necrosis and tumor in 66%.¹² In almost all cases, the tumor that is seen is of reduced viability.^20,21

Recurrence after Surgery

Surgery may fail because of anatomic considerations, pathologic features, or errors in judgment or technique. The involvement of critical structures may limit the initial resection: involvement of the optic pathways, diencephalon, internal capsule, brain stem, or eloquent cortex by a glioma often precludes complete removal. Tumor recurrence, despite removal of all macroscopically evident tumor, can occur if there is microscopic infiltration of adjacent structures. Even low grade cerebral gliomas are usually infiltrative. Anaplastic astrocytomas characteristically are widely invasive. Finally, errors in judgment, such as preoperatively underestimating the amount of tumor that can be safely removed or intraoperatively failing to remove tumor that was targeted, result in potentially resectable tumor being left as a nidus of regrowth.

Recurrence after Radiation

Radiation therapy may fail because of inadequate targeting, underutilization of tolerable dose, or radiation resistance of the tumor cells. Proximity to critical anatomical structures can also limit maximal allowable radiation dosage. Furthermore, the correlations between imaging abnormality and tumor extent are incomplete. Pathologic studies have shown that individual tumor cells can be found throughout and even beyond CT hypodense and MRI T2 hyperintense areas of malignant glioma.^23,24 For each individual case, the extent of brain invasion from the contrast-enhancing margin of a malignant glioma is incompletely known. The choice of field size for irradiation of such lesions is difficult and relies as much on the trade-off between target volume and tolerable dose as on the accurate delineation of tumor boundaries. Failure to include an adequate annulus of tissue about the tumor to accommodate imaging uncertainty and technical error may leave tumor cells incompletely irradiated.

Even if the maximal dose tolerated by infiltrative surrounding brain is delivered, tumor cells may remain viable. Hypoxic, nonproliferating cells and tumor stem cells are particularly radioresistant; with time or change in the physiologic conditions following therapy, re-entry of cells into the cell cycle permits the proliferation that results in clinically apparent tumor recurrence.^25,26 Analysis of patterns of failure demonstrates that, even with maximal tolerable doses of photon radiation of 70 to 80 Gy, almost all malignant astrocytomas fail centrally.²⁷ A study of high dose fractionated proton irradiation following radical resection of glioblastomas showed the following:¹ a dose between 80 and 90 Gy is sufficient to prevent tumor regrowth;² outside of this high dose volume, tumor regrows, usually in areas receiving between 60 and 70 Gy;³ enlargement of the high dose volume to include more peripheral areas is likely to induce unacceptably high levels of symptomatic radiation-related necrosis.²⁸

Recurrence after Chemotherapy

Chemotherapy fails as a result of inadequate drug delivery, toxicity, or cell resistance. The blood–brain barrier is deficient in the contrast-enhancing region of the tumor, but surrounding brain usually has an intact blood–brain barrier; lipid-insoluble drugs thus have limited access to tumor cells infiltrating peripheral regions. The margin between drug efficacy and neurotoxicity, bone marrow suppression, pulmonary injury, and gastrointestinal side-effects is often narrow. Non-cycling cells and tumor stem cells are resistant to cell-cycle specific drugs, and potentially vulnerable cells often rapidly develop biochemical means of resistance to chemotherapeutic agents.^25,26,29

Even if these therapies significantly reduce the tumor burden, the patient’s immune response may be rendered ineffective by chemotherapy and by the tumor’s secretion of factors antagonistic to immune function, such as IL-10, prostaglandin E2, and transforming growth factor (TGF)-beta 2, and its expression of apoptosis-inducing molecules, such as Fas ligand (FasL) and galectin-1.³⁰ Each of these limitations of each component of multimodality therapy may contributes to failure to prevent tumor regrowth. At the time of tumor recurrence, consideration of these reasons for failure is essential to assessment of prognosis and to the choice of subsequent therapy.

Residual Tumor

Radiologic demonstration of residual tumor after initial treatment may be consistent with preoperative goals and expectations; the prognosis would be that originally formulated. If, however, residual tumor is identified unexpectedly, the prognosis may need to be altered. The prognostic import of residual tumor is best seen in the relationship between extent of resection and likelihood of tumor recurrence.

Cytoreductive surgery is a fundamental part of the treatment of most systemic malignancies.³¹ In most cases, there is a strong relationship between the extent of resection and outcome. For gliomas, correlation between extent of resection or, more significantly, size of residual tumor, and outcome measures, such as interval to tumor progression and survival, has been strongly suggested by retrospective and prospective series, but not by randomized clinical trials.³² More recently, one study, reported a significant survival advantage when 98% or greater resection is achieved.³³

Correlation of survival with extent of resection for low grade gliomas has been suggested by retrospective uncontrolled reviews and comparisons with historical controls.^3,34–36 One study of 461 adult patients with low grade cerebral gliomas found that gross total surgical removal correlated with length of survival.³⁷ Another reported median survival duration of 7.4 years following maximal surgical resection. The median survival of a subgroup patients with hemispheric tumors compared favorably (10 years vs. 8 years) with that of a comparable series treated with biopsy and radiation alone.^3,35 Additional studies have demonstrated that extensive resection of low grade gliomas delays tumor recurrence.^36,38

For high grade gliomas, the correlations between the extent of resection at the initial operation and¹ the time to tumor recurrence and² the duration of patient survival have been disputed.³⁹ Historical reports and reviews of large series have noted the association of survival and extent of resection for both astrocytomas and oligodendrogliomas.^11,40–43 Extensive reviews of the literature, however, have failed to locate randomized, controlled clinical trials comparing survival after biopsy with that after radical resection of malignant gliomas.^44,45 Nevertheless, the benefit of surgical cytoreduction has been strongly suggested by the following findings:

Although less than ideal, the data that exist for gliomas and experience with tumors outside the central nervous system suggest the benefit of cytoreduction when a near-total removal (2 log reduction of tumor cell number) of a glial tumor can be achieved. Thus, failure to identify and remove readily accessible tumor mass at an initial operation might warrant reoperation before regrowth occurs.

Recurrent Tumor

Regrowth of tumors after an initial response (diminution or stability) to surgery and radiation therapy is ominous. This is particularly true if the growth is more rapid or more infiltrative than that of the original tumor. Such growth often exhibits changes in the basic biology of the tumor that make it less responsive to subsequent therapy. A short interval between initial treatment and recurrence of symptoms often indicates rapid regrowth and a poor prognosis. Factors to be considered in estimating prognosis include the biology of the tumor (its pathology, growth rate, and invasiveness), its resectability, its prior response to radiation and chemotherapy, and the age and performance status of the patient.⁵⁸ Estimates of the recurrent tumor’s size, growth rate, invasiveness, and location must be made in assessing its potential for causing both neurologic deficit and death. Reappearance of a slowly growing, well-demarcated frontal convexity oligodendroglioma with deletions of chromosomes 1p and 19q in a middle-aged patient of good neurologic condition after a 10-year interval of postsurgical quiescence clearly carries a prognosis very different from that of diffuse diencephalic spread of a glioblastoma multiforme in an elderly patient with a poor performance status 3 months after treatment with surgery, radiation, and chemotherapy.⁵⁹

Therapy of Recurrent Gliomas

The choice of therapy of a recurrent glioma is based on a comparison of the natural history of the regrowing tumor with the risk–benefit ratio of potential therapies. Recurrent gliomas warrant aggressive multimodality therapy if the patient is in good neurologic and general medical condition and therapeutic options offer a realistic chance for significant improvement in neurologic status or extension of survival.⁶⁰

Patterns of Recurrence of Gliomas

When gliomas recur, most do so locally. Historically, more than 80% of recurrent glioblastoma multiforme arose within 2 cm of the original margin of contrast-enhancing tumor.^61,62 In one series, over 90% of glioma cases showed recurrence at the original tumor location, while 5% developed multiple lesions after treatment.⁶³ In another study of 36 patients with malignant gliomas receiving 70 to 80 Gy of fractionated radiation, 32 (89%) had central (at least 95% of the recurrent tumor within the volume receiving at least 95% of the maximum dose) or in-field (at least 80% of tumor within this highest dose volume) recurrence, and 3 (8%) had marginal recurrence. Only one (3%) fell predominantly outside the high dose range. Seven patients had multiple sites of recurrence, but only one had a large recurrence outside the high dose volume.²⁷ This tendency to recur locally is a function of tumor cell distribution. There is a gradient of tumor cell density in which tumor cell number decreases rapidly at increasing distances from the contrast-enhancing rim of solid tumor. Thus, although individual tumor cells spread through the brain at great distances from the primary site, there are so many more cells locally that the odds favor local reaccumulation of tumor mass.^23,24

Factors contributing to the likelihood of local recurrence include the following:

As tumor cell proliferation resumes at the initial tumor site, cells again spread rapidly and diffusely. Tumor cell proliferation resumes at distant sites as a result of the influx of these new, mitotically active cells or the renewed growth of cells that spread before the initial treatment.⁶³ Biologic therapeutic agents may also affect the pattern of recurrence; recent experience with bevacizumab (Avastin) suggests that tumors treated with this inhibitor of VEGF are more likely to recur as diffusely infiltrative lesions distant from the original site of tumor.⁶⁴ Consequently, treatments targeting local recurrence alone will, at best, be briefly palliative. Treatment of tumor recurrence thus usually involves a combination of modalities aimed at both local and distant disease.

Epidemiology of Recurrent Glioblastoma

The heterogeneity in defining recurrence and the variability of treatment algorithms employed at different institutions result in a vague profile of recurrent glioblastoma multiforme.⁶⁵ In a multi-center trial of reoperation for resection and placement of cavitary biodegradable BCNU-wafer in 222 patients with recurrent glioblastoma and a Karnofsky Performance Score of at least 60, the median interval from initial diagnosis to tumor recurrence was 12 months.⁶⁶ Among a cohort of 301 patients with GBM, 223 patients had tumor recurrence at a median interval from initial diagnosis of 4.9 months;⁶⁷ 64% of these had a Karnovsky Performance Score greater than 70 at the time of recurrence.

Glioblastoma recurrence is demonstrated on imaging obtained in routine surveillance or in response to new or recurring symptoms. In a questionnaire-based study of patients with recurrent glioblastoma or anaplastic astrocytoma and a KPS >70, self-reported symptoms included fatigue, uncertainty about the future, motor difficulties, drowsiness, communication difficulties, and headache.⁶⁸ While most symptoms likely reflected tumor recurrence, confounding factors such as radiation necrosis and steroid treatment may have contributed to generalized fatigue, and pain and uncertainty about the future may have resulted from the diagnosis alone, independent of current tumor status. Incoordination, weakness, visual loss, and pain were reported more frequently by patients with recurrent glioblastoma than by those with anaplastic astrocytoma, providing evidence that more aggressive disease will cause greater neurological deficit.

Therapy of Recurrent Malignant Glioma

Choice of therapy for a recurrent glioma must consider the tumor’s current and previous histology, previous treatment, and location and the patient’s age, medical and neurologic conditions, and preferences. An enlarging lesion that was originally a low grade glioma should undergo biopsy (stereotactically or, if resection is anatomically feasible, by open craniotomy) to confirm histology (Fig. 10-2). If the tumor remains low grade and a large part of the lesion can be resected without inflicting significant neurologic deficit, it should be removed; if previously irradiated to significantly less than maximal tolerable dose, the tumor bed and surrounding area should receive fractionated radiotherapy. The longer the interval since the initial radiotherapy, the higher the dose that can be given safely at recurrence. If the tumor is inaccessible to surgery, radiation alone should be prescribed. If a low grade tumor previously irradiated to a maximal tolerable dose recurs as a low grade glioma, it should be resected, if possible. If it is inaccessible, stereotactically delivered focal radiation is an attractive option.^69,70

FIGURE 10-2 Recurrent glioma. Decisions in the management of a recurrent tumor should consider grade, resectability, and prior therapy. N, neurologic; reg rad, regional fractionated radiation therapy; focal rad/boost, stereotactic radiosurgery, brachytherapy, or radiotherapy; small, focal, less than 10 cm³, radiographically demarcated; CTx, chemotherapy; ST, stereostatic; XRTx, radiotherapy.

Biologic and Immune Therapies

Delineation of the molecular pathways of glial tumorigenesis has fostered development and testing of small molecule inhibitors and monoclonal antibodies targeting their components. Erlotinib and gefitinib, inhibitors of the epidermal growth factor receptor (EGFR)—a tyrosine kinase amplified or mutated in a high percentage of glioblastomas—are two examples of such molecular-based therapeutic agents. However, recent phase II trials have failed to demonstrate convincing effect. Gefitinib provided rates of progression-free survival at 8 months (PFS-6) of 14% in a prospective study of 28 patients with recurrent or progressive high-grade glioma and of 13% in 53 patients with recurrent glioblastoma.^111,112 A randomized, phase II trial conducted by the European Organization for Research and Treatment of Cancer found a PFS-6 of 12%, for patients with recurrent glioblastomas treated with gefitinib compared to a PFS-6 of 24% for the control group treated with either BCNU or TMZ.¹¹³ Disappointment over the failure of EGFR inhibitors as single agents has not dimmed optimism that they may be more valuable when used in conjunction with other therapeutic agents.

EGFRvIII, a constitutively active form of the receptor resulting from its most common deletion (found in each 30% of GBM), has been targeted by small inhibitory molecules and monoclonal antibodies specific for the mutated receptor. Tyrphostin, an inhibitor of EGFR more active against the mutated than against the wild-type receptor, has significantly delayed tumor recurrence in animal models and is entering the clinical phase of trials.¹¹⁴ And mAbs raised against the variant receptor have shown antitumor effects in cell culture.¹¹⁵

Since EGFR is an initial component of the PI3Kinase pathway, which is crucial to cell survival, proliferation and motility, tumor cells, when chronically activated by the EGFRvIII mutation, become dependent on PI3k, and thus potentially sensitized to its disruption.^116,117 Interestingly, the PTEN tumor–suppressor protein, an inhibitor of the P13K pathway, is often lost in glioblastoma.¹¹⁸ Based on these observations, it has been hypothesized that while possession of the EGFRvIII mutation would sensitize tumors to EGFR kinase inhibitors, loss of PTEN would mitigate this effect by disassociating EGFR inhibition from inhibition of the downstream P13K pathway. Coexpression of EGFRvIII and PTEN at both mRNA and protein levels was significantly associated with a clinical response in glioblastomas treated with EGFR kinase inhibitors.¹¹⁷ Thus, although inhibition of EGFR activity and the PI3K pathway may not be effective against all GBM, it may hold promise against tumors, at diagnosis or at recurrence, selected for having predisposing molecular lesions.

Targeting the tumor’s vasculature is another promising strategy.¹¹⁹ Bevacizumab (Avastin), an inhibitor of vascular endothelial growth factor (VEGF) with antiangiogenic and antiedema effects, rapidly reduces contrast enhancement at the site of tumor. This response is associated with increased time to tumor progression: bevacizumib and irinotecan produced a PFS-6 rate of 46% in patients with recurrent glioblastoma.¹²⁰ Despite other similarly impressive bevacizumib-induced increases in time to local recurrence of contrast-enhancing tumor, lengthened overall survival has not been observed. This suggests that the rapid, often substantial, reduction of contrast enhancement induced by bevacizumab may reflect decreases in vascular permeability rather than true regression of tumor. Furthermore, there is increasing concern both that the suppression of angiogenesis during treatment is transient and that treatment may favor more diffuse, initially hypoangiogenic growth.⁶⁴ Despite these concerns, an advisory committee to the Food and Drug Administration recently unanimously recommended approval of bevacizumib for treatment of glioblastoma.

Immunotherapy also holds great promise for patients with malignant gliomas.¹²¹ Efforts to enhance the immune response to tumors include both passive and active strategies.¹²² Passive approaches include implantation of modified immune cells into a resection cavity. One early study employing lymphokine-activated killer (LAK) cells and IL-2 reported a median duration of survival of 53 weeks after reoperation and implantation compared with 26 weeks following reoperation and chemotherapy.¹²³ Another implanting the LAK cells without the IL-2 in 40 patients with recurrent glioblastomas reported a median duration of survival of 9 months and a 1-year survival rate of 34%.¹²⁴ Adoptive immunotherapy, a variety of passive immunotherapy, is receiving great interest. Tissue harvested at surgery is disaggregated and various constituents (simple lysate, DNA, proteins, etc.) are used to prime dendritic cells harvested from the patient’s blood or tumor.^125,126 Active immune strategies employ vaccination. Clinical testing of a vaccine to EGFR vIII has progressed to a phase III trial based on safety and efficacy against newly diagnosed GBM in phase I and II trials.¹²⁷ Confirmation of the actual clinical benefit of these and other efforts to enhance the immune response awaits completion of phase III trials.

Monoclonal antibodies to other tumor cell surface receptors have also been used to block other components of growth factor signaling pathways (e.g., PDGF), to deliver toxins specifically to tumor cells (e.g., toxin linked to EGF or IL-13), and, as mentioned above, to selectively irradiate tumor sites (e.g., I131-mAb).^90,128 For example, Pseudomonas exotoxin has been targeted to the IL-13 receptors highly expressed on malignant glioma cells by injecting the toxin conjugated to IL-13 into the resection cavity of recurrent, malignant glioma. Pseudomonas exotoxin conjugated to TGFalpha also has been investigated in a phase I trial, with relative safety but mixed results in terms of efficacy. Overall, toxin delivery by cytokines has demonstrated relative safety and variable efficacy.

Cells carrying therapeutic genes have also been used. Preliminary gene therapy studies using ganciclovir activated by a thymidine kinase gene delivered by a retrovirus produced by a modified mouse fibroblast packaging cell line have proved feasibility and safety but have not yet achieved proof of mechanism or demonstrated efficacy.^129,130 In a small study that examined the effects of an intratumoral injection of retroviral vector-producing cells combined with intravenous ganciclovir, the authors obtained a 1-year survival rate of 25% with tumor response in 50% of the cases.¹³¹

Numerous such strategies utilize reoperation to harvest tissue for molecular analysis or to guide tumor specific therapies and as a source of agents for vaccination. Reoperation also presents an opportunity to implant drug polymers, immune stimulants, toxins, viruses, radioisotopes, therapeutic cells, and infusion catheters for subsequent delivery of therapeutic agents (Fig. 10-3).^66,123, ¹²⁸, ¹²⁹, ¹³¹

FIGURE 10-3 Adjuvant therapy. Reoperation provides the opportunity to implant drug polymers, immune stimulants, toxins, viruses, radioisotpoes, therapeutic cells, and infusion catheters for subsequent delivery of therapeutic agents. A 39-year-old man developed marginal recurrence about the cystic resection cavity left after removal and irradiation of an anaplastic astrocytoma 3 years previously. As part of a gene-therapy trial, three columns of vector-producing cells were infused within the tumor, at the tumor margin, and in the surrounding tumor-infiltrated brain. A, Coronal MRI. Five days later, the tumor and the infused tissue to be analyzed were removed. Care was taken to extend the resection to the margin of eloquent areas and to achieve watertight closure. B, Coronal, immediately postoperative MRI.

Rationale for Reoperation

Early reoperation, within months of the initial procedure, might be indicated for complications such as intracerebral, subdural, or epidural hematoma, wound dehiscence and infection, or hydrocephalus and CSF leakage. Occasionally, failure to identify and remove readily accessible tumor mass might warrant reoperation. In the Royal Melbourne Hospital experience, 5 of 200 patients underwent early reoperation.¹³² More frequently, true tumor recurrence after an interval of response to the initial therapy is the reason for considering reoperation. Reoperation is justified if it produces sustained improvement of neurologic condition and quality of life or significant increase of response rates to adjuvant therapy. Palliation of neurologic symptoms by surgery results from reduction of the local mass effect produced by the tumor and tumor-induced edema. Steroid dose may be able to be reduced and steroid side effects diminished.

Multiple studies have shown that initial surgical cytoreduction of a malignant glioma can both improve neurologic deficits and promote maintenance of high-performance status. One review of 82 patients examined five categories of neurologic function in each patient. Preoperatively, 191 neurologic deficits were noted. Postoperatively, 151 deficits were improved or stable and 40 were worse.¹³³ Another study showed that patients undergoing gross total resection of their malignant gliomas were likely to have improved neurologic function (97% of 36 patients had either improved or stable neurologic examinations), higher functional status (mean KPS score improvement of 6.8%), and extended maintenance of good functional status (185 weeks).^34,51 A third study confirmed that more extensive resection was correlated with better immediate postoperative performance, lower 1-month mortality rate, and longer survival: 43% of patients with malignant gliomas improved, 50% remained unchanged, and 7% suffered deterioration in their neurologic condition following resection of at least 75% of their tumor. A more limited resection proved inferior (28% improved, 51% were unchanged, and 21% were worse).⁴⁹

Similar results can be achieved by reoperation. In one series, 45% of patients had an improved Karnofsky score following reoperation.¹⁰⁸ In another series focusing on reoperation, when gross tumor resection was achieved, 82%^32,39 of patients had improvement or stability in their Karnofsky score.¹³⁴ In a third, patients with Karnofsky scores of 50 or less also underwent reoperation. Two-thirds improved from a dependent to an independent state and the median duration of survival was similar to that for all patients undergoing reoperation.¹³⁵

The doubling rate of malignant gliomas is so high, however, that the benefit from reoperation will be very brief unless adjuvant therapies are able to induce remission of tumor growth. Surgical resection is especially beneficial when reduction of tumor burden improves the response rate to such adjuvant therapies. One early study of multidrug chemotherapy following reoperation for recurrent malignant gliomas used multivariate analysis to identify prognostic factors. During chemotherapy, disease stabilization or partial response occurred in 29 of 51 (57%) patients. Median time to tumor progression was 19 weeks for all pathologies; it was 32 weeks for patients with anaplastic astrocytomas and 13 weeks for those with glioblastoma multiforme. Median survival time was 40 weeks for all pathologies; it was 79 weeks for patients with anaplastic astrocytoma and 33 weeks for those with glioblastoma multiforme. Thirty-five percent of patients had serious chemotoxicity, but none had permanent morbidity or mortality. Factors associated with a longer median time to tumor progression included a higher Karnofsky score, lower grade of initial histology, lack of prior chemotherapy, less myelotoxicity, smaller postoperative tumor volume, greater extent of resection, and a local rather than diffuse pattern of recurrence. Those associated with a longer median survival time were a higher Karnofsky score, anaplastic astrocytoma rather than glioblastoma at recurrence, greater and myelotoxicity, and lobar rather than central location of the tumor.⁷²

More recent studies from the University of California-San Francisco (UCSF), Memorial Sloan-Kettering, the University of Washington, and Johns Hopkins University have shown that reoperation followed by chemotherapy leads to stabilization of performance score for significant intervals.^{66,108,136,137} At UCSF, 44% of patients with glioblastomas maintained a performance level of at least 70 (a level consistent with self care and judged to be survival of high quality (58%)) for at least 6 months after reoperation; 18% maintained this level for at least a year; and 3 patients did so for longer than 3 years. Most (52% of 31) patients with anaplastic astrocytomas maintained this performance level for at least 12 months after reoperation; 13% had more than 4 years of high quality survival. Approximately 90% of the survival after reoperation for anaplastic astrocytoma was of high quality¹³⁶ (Fig. 10-4). In the Memorial Sloan-Kettering group, the median duration of maintenance of independent status (Karnofsky score of at least 80) was 34 weeks. In the Seattle series, patients with a Karnofsky score of at least 70 maintained this high level of function for an average of 37 weeks after reoperation for glioblastoma and for 70 weeks after reoperation for anaplastic astrocytoma.¹³⁷

FIGURE 10-4 Quality-of-life considerations in the management of patients with malignant gliomas. Maintenance of high-performance status is a critical feature of outcome. The Karnofsky score (KPS) as a function of time indicates the quality of survival time that follows each intervention. 1S, initial surgery; XRT, radiation therapy; CMT, chemotherapy; 2S, reoperation; TS, total survival; HQS, high-quality survival (K = ≥70); SAR, survival after reoperation; AA, anaplastic astrocytoma; GBM, glioblastoma multiforme.

A more complete resection of recurrent tumor increases the duration as well as the quality of patient survival.^60,134 Support for reoperation is found in comparisons of the outcomes in cases in which different degrees of tumor removal were accomplished and in comparisons of the survival of patients following reoperation with that of historical controls not receiving reoperation. Patients in whom gross total resection of a glioblastoma is achieved survive longer (45.6 vs. 25.6 weeks) than do those receiving near-total or subtotal resections; for anaplastic astrocytomas, the effect of extent of resection is similar (87.5 vs. 55.7 weeks).¹³⁷ Comparable results were obtained by another group in which the median duration of survival for GBM patients after gross total resection was 76 weeks and for anaplastic astrocytoma was 33 months.¹³⁸ In the Sloan-Kettering series that grouped glioblastomas and anaplastic astrocytomas together, a similar difference was found (51.2 vs. 23.3 weeks).¹⁰⁸ In the UCSF series, survival of patients undergoing reoperation and chemotherapy for either anaplastic astrocytoma or glioblastoma was longer than that of patients receiving chemotherapy alone at the time of tumor recurrence.¹³⁶

A later study from UCSF⁶⁷ evaluated 46 patients (15.3%) who underwent reoperation among a group of 301 patients with glioblastoma multiforme. The actuarial rate of reoperation was 15% at 1 year and 31% at 2 years after the initial diagnosis. Patients who were younger (p = 0.01) and who had received an extensive resection initially were more likely to undergo reoperation. KPS scores after reoperation were improved in 28% of patients, unchanged in 49%, and worse in 23%. There was no perioperative mortality. Eight (17.4%) underwent a third operation. A high preoperative KPS score was the only factor predictive of longer survival after reoperation. Similar to the earlier study, the median duration of survival after reoperation was 36 weeks: 61% were alive at 6 months and 24% at 21 months after reoperation. The median period of high-quality survival was 18 weeks (Table 10-1). A subset of 32 patients who underwent reoperation within 45 days of first tumor recurrence was compared with a control group of 141 patients who did not. The median duration of survival after recurrence was 18.7 weeks longer (42.4 vs. 23.7 weeks) in the group undergoing reoperation (p < 0.05, hazard ratio = 0.67, 95% confidence interval = 0.44–1.00) even when controlled for age and KPS score at the time of recurrence. The 13-week difference in survival was potentially due to selection bias. Adjusting for resectability by eliminating from the control group those who underwent biopsy only as the first operation and those whose recurrent tumor was separate from the original tumor reduced the statistical significance of this difference (p = 0.12; hazard ratio = 0.71, 95% confidence interval = 0.46–1.09). The Cox multivariate proportional hazards model adjusted for resectability predicted that a typical 55-year-old man with a KPS score of 80 likely would survive 8 weeks longer (35 vs. 27 weeks) with reoperation than without it. Selection bias was also reduced by stratifying reoperation and control groups by propensity to undergo reoperation. This analysis identified reoperation (p = 0.03, hazard ratio = 0.64, 95% confidence interval = 0.42–0.96) as well as KPS score as statistically significant predictors of longer survival.⁶⁷

TABLE 10-1 Karnofsky Performance Status

From Karnofsky D, Burchenal JH, Armistead GC Jr, et al. Triethylene melamine in the treatment of neoplastic disease. Arch Intern Med. 1951;87:477-516.

The benefit of reoperative surgery is also suggested by experience with brachytherapy. Patients undergoing reoperation for tumor recurrence and/or radiation necrosis following brachytherapy for glioblastomas, either initially or at first recurrence, survived longer than those not receiving reoperation (median total survival of 120 vs. 62 weeks for patients with primarily treated tumors and 90 vs. 37 weeks for patients treated with brachytherapy at the time of first recurrence).¹²

Reoperation as a part of the multimodality treatment of recurrent gliomas is further supported by study of long-term survivors of glioblastoma multiforme. A review of the UCSF experience identified 22 of 449 (5%) patients with glioblastomas who survived at least 5 years after diagnosis. Sixteen of 22 had tumor recurrence that was treated; and 9 underwent between one and three reoperations. For 8 of the 16 patients with treated recurrence, survival after treatment of recurrence (median of 4.5 years) was longer than the remission produced by the initial treatment.¹³⁸

The benefit to survival from reoperation is not above dispute. As noted above, multivariate analyses of chemotherapy studies have found that extent of resection and smaller postoperative volume are associated with prolongation of time to tumor progression but not of survival.³² A similar study of survival after progression of malignant gliomas identified high KPS score and age less than 50 years as independent prognostic factors.¹³⁹ Those who underwent reoperation (58/143 patients) lived longer (median of 35 weeks vs. 16 weeks, p < 0.005 in univariate analysis) after tumor recurrence than those treated without reoperation, but multivariate analysis identified only a trend toward reduction of risk of death (hazard ratio = 0.74; 95% confidence interval = 0.50–1.11; p = 0.014) following reoperation. Randomized controlled trials using prognostic indices for analysis of outcomes are needed to evaluate definitively the benefits of reoperation.¹⁴⁰

Selection of Patients for Reoperation

Case selection is critical to outcome. The patient’s profile of prognostic factors, the predicted tolerance of the procedure, and the feasibility of extensive tumor resection without undue risk of new neurologic morbidity must all be considered. Multiple characteristics have been identified as predictive of a good response to reoperation (Table 10-2). Foremost among these are tumor histology, patient age, performance status, interoperative interval, and extent of resection.

The prognostic significance of tumor grade is evident in most series. Median duration of survival after reoperation was 88 weeks for patients with anaplastic astrocytomas but only 36 weeks for those with glioblastomas at UCSF, and 61 weeks and 29 weeks, respectively, at Sloan-Kettering.^108,136

The effect of age may overwhelm that of tumor grade. In one series, survival after reoperation was 57 weeks for those younger than 40 years but only 36 weeks for older patients.⁶⁰ Other authors found an association between youth and total survival duration after diagnosis and between youth and quality of survival after reoperation, but not between youth and duration of survival after reoperation.^108,136

Preoperative performance score significantly affects outcome. At Kentucky, median duration of survival after reoperation was 22 weeks for patients with performance scores of at least 70 but only 9 weeks for more disabled patients.¹⁴¹ In the Seattle series, for glioblastomas, survival after reoperation was almost twice as long (71 vs. 36 weeks) for patients with Karnofsky scores of at least 70 than for those with lower scores.¹³⁷

The prognostic importance of the interval between initial treatment and recurrence is disputed.¹⁴² The Kentucky series found survival to be twice as long if the interval between operations exceeded 6 months. In Seattle, a threefold difference (150 vs. 48 weeks for glioblastoma, and 164 vs. 52 weeks for anaplastic astrocytomas) was noted when the time to progression exceeded 3 years.¹³⁷ Others, however, have found either no relation, or an inverse relation, between the interoperative interval and duration of survival after reoperation.^60,108,136

The likelihood of substantial resection must also be anticipated. The ability to remove sufficient tumor mass to reduce intracranial pressure and palliate neurologic symptoms depends on the location of the tumor and its physical characteristics. Removal is facilitated by a more superficial location in noneloquent areas, a discrete pseudoencapsulated mass is more easily removed than a less well-marginated diffuse one, drainage of a cystic component will often provide immediate reduction of mass, as well as an avenue for further tumor resection.

During the interval between initial surgery and tumor recurrence, the patient will usually undergo therapy that might affect tolerance of surgery. The decision to reoperate must consider overall physical condition, tissue viability, blood coagulability, hematologic reserve, and immune function following surgery, radiation, corticosteroids, and chemotherapy. A high risk of multisystem failure, failure to thrive, intracranial hemorrhage, anemia, wound infection, pneumonia, and neurologic damage may exist. This risk should be assessed for each patient by obtaining preoperative chemical, hematologic, and radiographic studies.

In choosing patients for reoperation and subsequent therapy, consideration of the individual patient’s profile of these prognostically significant factors permits a reasonable estimate of the likelihood of benefit from the procedure. Quality of life is a very important determinant of benefit as perceived by the patient, family, and society. Reoperation and subsequent multimodal therapy should be chosen only if maintenance of a reasonable level of function is anticipated (Figure 10-4).

Preparation for Reoperation

Before surgery, the patient is likely to be receiving corticosteroids; these should be continued. At the time of induction of anesthesia, the patient is fitted with thigh-high intermittent-compression air boots. Additional steroids, prophylactic antibiotics, and osmotic and loop diuretics are given and the patient is hyperventilated. In positioning the patient, the likelihood of elevated intracranial pressure makes attention to elevation of the head above the level of the heart particularly important.

Reoperative Exposure

In planning the needed exposure, the tumor can be located by its relationship to the margins of the craniotomy and to the cortical pattern of gyri and sulci on the preoperative MRI scan, by intraoperative frameless stereotaxy, or by intraoperative MRI.^143,144 The procedure should be planned in advance to ensure adequate skin opening, craniotomy, and durotomy to expose the recurrent mass. All may need enlarging because of the increased extent of the tumor or a desire to perform corticography for mapping of motor or speech function.

The skin incision from the previous operation is usually used. The skin opening can be increased by additional incisions. They should be external to the previous flap, avoid its base and other vascular pedicles, and intersect the previous incision at right angles. The margins of the prior craniotomy flap should be defined. Generally, this is best accomplished with a curette, beginning at the prior trephination. Depending on the interoperative interval, the prior flap may need to be recut. Dissection in the epidural plane can be begun with a curette followed by a No. 3 and then a No. 2 Penfield dissector. The craniotomy plate is further elevated as the dura is stripped from its inner surface with a dissector. Epidural adhesions fixing dura to the craniotomy margin should be preserved as prophylaxis against postoperative extension of an epidural fluid collection unless the craniotomy needs to be enlarged. In this case, these adhesions are dissected with a curette and trimmed. After the dura is stripped from the undersurface of the cranial plate, an additional segment of bone can be removed with a craniotome.

The durotomy may need to be enlarged, but often it can be limited to part of the dural exposure. It should be planned to minimize traverse of cortical adhesions. For instance, in re-exposing a temporal lesion, the durotomy can be placed over the cyst remaining from the prior resection. Flapping the dura superiorly then allows adhesions to be put on traction such that they may be dissected from cortex, coagulated, and sharply divided. Extending a durotomy along an old incision line should be avoided. The old incision line should be traversed perpendicularly and as infrequently as possible in that it is often the site of the densest adhesions. Microdissection of larger vessels from dural attachments may be necessary. Once the dura is opened and retracted, the exposed cortex is inspected for the surface presentation of the tumor, apparent as abnormal color, consistency, and vascularity.

Localization of the subcortical extent of the tumor is then undertaken. Again, the preoperative imaging studies and stereotactic techniques are of value.¹⁴⁵ Brain shift occurring during dural opening and tumor resection must be considered when using these techniques. Intraoperative MRI or ultrasonography may help correct for this.^144,146,147 Tumor may also be found by locating a cystic resection cavity or encephalomalacic brain left after the previous operation. In that almost all tumors recur within 2 cm of the original tumor’s margin, exposure of the initial tumor’s surgical bed will usually reveal at least part of the recurrent mass.

Electrocorticographic mapping of motor, sensory, and speech areas may reduce the chance of inflicting neurologic deficit and may encourage a more extensive resection by revealing the relationship of the site of cortical traverse and of the subsequent subcortical dissection to eloquent brain (Fig. 10-5). This technique is often more difficult at the time of reoperation because of cortical disruption by the tumor and prior surgery.¹⁴⁸ However, long term reshapings of language, sensory, and motor maps in some patients have allowed gross total resection of recurrent gliomas without neurologic sequelae.^149,150

FIGURE 10-5 Recurrent malignant glioma. A 43-year-old woman developed right arm and leg weakness 11 months following complete resection and irradiation (60 Gy) of a right frontal glioblastoma. Pre-reoperative axial (A) and sagittal (C) MR scans showing contrast enhancement just posterior to the resection cavity. Reoperation, guided by intraoperative electrocorticographic mapping of the primary motor area, accomplished gross total removal of the recurrent tumor and surrounding frontal lobe back to the precentral sulcus, as seen on postoperative axial (B) and sagittal (D) images. After resolution of perioperative edema, full extremity strength returned.

Generally, the appearance of the tumor itself is the best guide to its extent. Tumor-infiltrated cortex is likely to have increased vascular markings, a pink to gray color, and a firm consistency. Its central core may vary from yellow cystic fluid of low viscosity to high viscosity, soupy, white necrosis which resembles pus, to a yellow-gray, granular, honeycomb-like material. Typically, the center is relatively avascular although it may be traversed by thrombosed blood vessels.

Some advocate incision into the tumor mass and internal debulking with an ultrasonic aspirator or laser as an initial step; however, this often induces significant hemorrhage. Enucleation by circumferential dissection in the pseudoplane about the rim of solid tumor is usually more satisfactory. Arteries supplying the tumor and veins draining it can be coagulated and divided as they enter the tumor mass, much as the vascular supply of an arteriovenous malformation is handled. In areas of noneloquent brain in particular, the softened, necrotic, highly edematous white matter around the tumor provides an excellent plane of dissection. The use of bipolar cautery forceps and suction together accomplishes this dissection while reducing local mass. Beyond the encephalomalacic brain lies more normal brain which, although edematous and possibly injured by prior retraction and radiation therapy, is often functional and should be preserved.

Often the tumor can be removed as a single specimen without the need for significant retraction of surrounding brain. In general, gentle, transient gentle pressure on a cottonoid paddy lying on the margin of resection provides sufficient exposure of the dissection plane that fixed self-retaining retractors are unnecessary. Retraction of the tumor mass is preferable to retraction of surrounding brain. Often, identification of the appropriate plane for the circumferential dissection is facilitated by this retraction on the tumor; coherence of the tumor mass helps delineate the plane between solid tumor and tumor-infiltrated brain.

Once the tumor mass has been removed, the margins of resection should be inspected to verify completeness of the excision. The margins should be free of tumor that is more firm, glassy, opaque, and hypervascular. Biopsies of the surrounding edematous brain should be sent for frozen-section analysis to verify absence of tumor. If solid tumor or tumor infiltrating into noneloquent areas remains, it should be removed. In some cases, extension of tumor into eloquent areas or diencephalic structures will preclude resection of the entire mass. In such cases, the tumor should be divided. This often entails coagulation of numerous strands of small, thin-walled blood vessels. This is particularly true if the extension is in the direction of the vascular supply, such as medial extension of a temporal lobe tumor toward the posterior aspect of the sylvian fissure. Particular care should be taken to coagulate and sharply divide these vessels. Tearing them without prior coagulation will leave a loose end that will retract and continue to bleed. Such loose ends should be directly coagulated rather than tamponaded with hemostatic packing, which may encourage deeper dissection of a hematoma.

After the resection has been completed, hemostasis should be confirmed by filling the tumor cavity with saline and, during a Valsalva maneuver, observing for wisps of continuing hemorrhage. This should be performed with the patient’s blood pressure at least as high as the normal pressure. The cavity is then aspirated, lined with a single layer of oxidized cellulose, and filled again with irrigation fluid. Hyperventilation is then reversed to permit expansion of the brain during closure.

Watertight dural closure is essential. Often, this can be attained by primary suturing, given the decompression from the operation. If the dura is incompetent, it may be supplemented by a pericranial graft or a synthetic dural replacement. Peripheral and central dural tacking sutures are placed. The bone fragments are wired together, and the craniotomy plate is then fixed with titanium miniplates. The wound is irrigated repeatedly with antibiotic solution and then closed in layers with 2-0 absorbable suture in muscle, fascia, and galea. The galeal sutures should be inverted and the knots should be cut short to avoid erosion superficially. They should be placed in sufficient proximity that tension-free closure of the skin is possible. Staples or simple running 4-0 nylon skin sutures usually provide adequate closure except at sites of attenuation, where horizontal mattress sutures may be preferred.

Postoperatively, the patient should be closely monitored for signs of increased intracranial pressure from hematoma or edema. Fluid restriction, dehydration, and corticosteroids should usually be continued. The patient should be mobilized as soon as possible and a gadolinium enhanced MRI scan should be obtained as soon as it is tolerable to assess extent of tumor resection.