Anaplastic astrocytoma (AA) defined by the International Classification of Disease for Oncology (ICD-O) code as 9401/3 (ICD-O 2000) is uncommon in adults. In Europe, the annual incidence is about 0.3 per 100,000 (EUROCARE 2004). In population-based registries, AA constitutes 4% of all malignant (EUROCARE) nervous system tumours (ICD-O 191,192) (ICD-O 1990). In Europe, about 60% of patients with a diagnosis of AA are between 45 and 69 years of age. In this age group the annual incidence rate ranges from 0.5 to 0.7 per 100,000 (EUROCARE 2004). AA is 1.4 times more common in men (EUROCARE 2004). A study on the trends in incidence of adult primary intracerebral tumours in Denmark, Finland, Norway, and Sweden found an increase in the overall incidence during 1969-98 that was confined to the late 1970s and early 1980s (Lonn 2004). Since 1984, the incidence has been stable or has even shown a minor decrease. In the analyses of specific histological types during the period 1969-98, the incidence of astrocytoma remained constant.

Survival data for AA is available from population-based cancer registries of 18 European countries in the EUROCARE study (Berrino 2003; Roazzi 2003). The survival analysis covered 1,064 adults with a diagnosis of AA during the period 1990-94 and followed up until 1999. Prognosis for AA is poor. Relative survival for adults diagnosed with AA during 1990-1994 was 44% at one year, 22% at three years, and 16% at five years, showing no difference between men and women. Five-year relative survival decreased markedly with age from 33% in the youngest (15-45 years) age group, to 2% in the oldest age group of patients (65 years and over). There have been significant improvements in survival since the early 1980s. In Europe, during the period 1983-94, one year survival rose from 26% to 43%, and the five-year survival improved from 9% to 16% (Roazzi 2003).

The causes of astrocytoma are largely unknown, but genetic factors and a variety of environmental factors have been implicated at different levels.
Certain inherited syndromes, such as neurofibromatosis type 1 and 2, tuberous sclerosis, and Li-Fraumeni syndrome, predispose to developing astrocytomas. Mutations in p53 are found in two-thirds of patients affected by low grade diffuse astrocytomas (WHO grade II) (Chawengchao 2001). Anaplastic astrocytoma (WHO grade III) is accompanied by genetic alterations, including loss of heterozygosity on chromosome 19 (Preston-Martin 1996).
Many studies of familial risk for astrocytoma among adults were conducted through the nationwide Swedish Family-Cancer Database. Among adult offspring, close to 11,000 patients with CNS tumours were identified, out of whom about 200 had a parent diagnosed with a CNS tumour. The standardised incidence ratio (SIR) was double for astrocytoma when parents were diagnosed with astrocytoma, while among siblings the ratio was tripled (Hemminki 2003a). Parental endometrial cancer and melanoma were associated with astrocytoma in their offspring (Hemminki 2001) and a familial risk of over 4 was found for low-grade astrocytoma (Hemminki 2003b).
In a case-control study, conducted in Los Angeles county, a detailed job history and information about other suspected risk factors were obtained from about 300 adult patients with CNS tumours and 300 controls (Preston-Martin 1989). A positive association was discovered between gliomas and full-mouth dental X-ray examinations after age 25, employment in jobs likely to involve high exposure to electric and magnetic fields, and more cases had worked in the rubber industry and in hot processes using plastic.
Aschengrau and colleagues found an association between astrocytomas and residential proximity to cranberry bog cultivation in a population-based case-control study conducted among residents in the Massachusetts area (Aschengrau 1996). Several occupational studies showed high risk of astrocytomas in people with electrical and electronics jobs (Thomas 1987a). Risk was higher for engineers, teachers, technicians, repairers, and assemblers. Risk increased with duration of exposure to tenfold among those employed for 20 or more years.
An increased risk of astrocytoma was found by a case-referent study conducted in northern New Jersey for workers exposed to organic chemicals in petroleum refining and chemical manufacturing (odds ratio between 1.4 and 1.8) (Thomas 1987b).
An association of astrocytic brain cancer with exposure to carbon tetrachloride, methylene chloride, tetrachloroethylene, and trichloroethylene, was found, which was strongest for methylene chloride (Heineman 1994).
From a Swedish study based on the Swedish Cancer Enviroment Registry, pulp mill workers were found to have more gliomas than expected (SIR 1.5) (Andersson 2002).
A case-control study of gliomas diagnosed among adults in the San Francisco Bay area assessed the occupational risk factors and confirmed previously reported associations with several occupations. Legal and social service workers, shippers, janitors, motor vehicle operators, and aircraft operators had increased odds ratios, but only in those people with longer duration of employment. Physicians and surgeons, foundry and smelter workers, petroleum and gas workers, and painters showed increased risk for both astrocytic and nonastrocytic tumors (Carozza 2000). From a similar study, increased risk of gliomas was associated with several other occupations: butchers and meat cutters, computer programmers and analysts, electricians, general farmers and farmworkers, inspectors, checkers, examiners, graders, and testers, investigators, examiners, adjustors, and appraisers, physicians and physician assistants, and store managers, whereas employment as a childcare worker was associated with decreased incidence of glioma (De Ross 2003).
A Swedish population-based case-control study on the use of cellular and cordless phones found a doubling of the risk for astrocytoma with the ipsilateral use of an analogue cellular phone (Hardell 2002).
Population-based case-controlled data collected in eight Canadian provinces did not support the hypothesis that occupational magnetic field exposure increases the risk of astrocytoma (Villeneuve 2002).

Acknowledgment
The authors thank the members of the EUROCARE Working Group for their permission to use the survival analysis from the EUROCARE dataset.

Astrocytomas are currently best classified according to the revised WHO classification (Kleihues 2000; Kleihues 2002). While grade I and II are considered low-grade gliomas, grade III (anaplastic astrocytoma) and grade IV (glioblastoma multiforme) are commonly referred to as malignant gliomas. Tumours may evolve over years from low grade to a higher grade or may develop de novo as a malignant glioma. Within a given tumour, differences in grade may be observed. However it is the most malignant grade that defines the diagnosis and will determine the prognosis. Anaplastic astrocytomas frequently evolve from a less malignant precursor lesion and may further transform into glioblastomas. In secondary malignant gliomas, p53 mutations and PDGFR overexpression are commonly found. While there is usually a good concordance in diagnosis for grade IV tumours between pathologists, disagreement in >30% of samples is common for anaplastic astrocytoma and other grade III tumours.

Depending on localization, tumour size and associated brain oedema, patients will present with variable neurological symptoms. Seizures, blurred vision, focal deficits or progressive cognitive deterioration are common features at presentation. Brain imaging with computer-assisted tomography (CT) or magnetic resonance imaging (MRI) will show a localized expansive process with associated oedema. Brain lesions are often irregular in shape and contrast-enhancing, frequently with a necrotic centre. Differential diagnosis includes abscess or metastatic brain tumour. Primary CNS lymphoma and other primary brain tumours may present in a similar way. Although calcifications, or the intensity of contrast enhancement, may be indirect signs of tumour histology (e.g. calcifications in oligodendroglioma, and contrast enhancement with a necrotic core in glioblastoma), these criteria are highly unspecific. Subsequent histological confirmation by stereotactic biopsy or after tumour resection are indispensable for diagnosis. One drawback of stereotactic biopsies is that a few millimeters of tumour material is not always representative of a much larger tumour. In order to establish a precise diagnosis, care should be taken to provide the pathologist and ideally the molecular diagnostics laboratory with a sufficient amount of tumour tissue. Biopsy should be taken from the contrast-enhancing margin of the lesion rather than the necrotic core.

Unlike most other cancer types, malignant astrocytomas do not metastasize and tend to present as a localized brain lesion. At a late stage in occasional long-term survivors diffuse dissemination in the brain and in the cerebro-spinal fluid may be observed. Distant metastases are a rare and exceptional event, possibly associated with contamination during suregry. Thus no formal staging outside the brain is required. Preferred imaging in the brain is gadolinium-enhanced magnetic resonance (MRI). For patients undergoing complete or partial tumour resection, an immediate, (preferably within the first 24 hours to maximum 48 hours postoperative) MRI will allow the extent of residual tumour to be determined. Imaging after a longer interval will show post-surgery contrast-enhancement which cannot be distinguished from tumour tissue.

Only few data exist and specifically address management of anaplastic astrocytoma. Until recently most trials pooled both grade III and grade IV astrocytoma as malignant glioma. Therefore some of the conclusions are based on extrapolation and rational interference from high-grade glioma in general, on a type R basis. However, recent progress in molecular diagnostics, biology and understanding of glioma have clearly demonstrated that different disease entities exist on a molecular level and in clinical outcome. Thus, most ongoing trials will separate grade III and grade IV astrocytoma, and also oligodendroglioma and oligoastrocytoma with LOH 1p/19q.

The gold standard in proving or disproving the efficacy of a therapeutic intervention is a large phase III trial demonstrating a clear survival benefit. However, in rapidly growing and devastating diseases like anaplastic gliomas, the goal of therapeutic efforts is also palliation of symptoms and maintenance of quality of life. Comparison between reports for therapy of malignant glioma is subject to a multitude of pitfalls. Firstly, most reports in patients with recurrent disease comprise an amalgam of varying histologies, including grade III and grade IV astrocytoma, transformed low-grade glioma as well as oligoastrocytoma and oligodendroglioma. Many phase III trials of the 1980's appear underpowered to detect a difference between treatment arms. Further, histologies are mixed and not necessarily comparable to today's classification. Of note also is the large percentage of ineligible patients. In phase II trials surrogate endpoints such as response rate are frequently used in medical oncology. However, measuring contrast enhancement on imaging has its inherent difficulties, due to the infiltrative nature, and non-homogenous cellular composition with areas of necrosis. This leads to substantial interobserver variability (Vos 2003). Furthermore, the use of corticosteroids may change the appearance of tumour on CT or MRI. Over 10 years ago, Macdonald and coworkers suggested also including neurological function and steroid use in the assessment of tumour response for gliomas (Macdonald 1990). Also, responses in the brain may be delayed and are occasionally only seen many months after treatment starts. Objective response rates in recurrent anaplastic astrocytoma are usually between 10-30% and in one recent trial up to 35% (Yung 1999). As progression is easier to identify than measuring tumour shrinkage, progression-free survival at 6 months (PFS6mo) has been suggested as a surrogate endpoint. Based on the pooled results of 8 consecutive and negative phase II trials in patients with grade III and grade IV astrocytoma this new endpoint has been proposed (Wong 1999). The authors suggested considering a PFS at 6 months of 31% for anaplastic astrocytoma as a benchmark for comparison in phase II trials (Wong 1999). Most current trials of new chemotherapy agents for recurrent glioma refer to this endpoint, however, validation in a prospective randomized trial or by an independent dataset is still lacking.

Surgery serves both the purpose of gaining tumor tissue for exact histological diagnosis and for reducing the tumour bulk. Although a complete resection is usually attempted, as far as it is safe and feasible, the true value of this approach has never been investigated in a prospective controlled trial and it is, therefore, to be considered as suitable for individual clinical use on a type C basis. Some conclusions can be drawn form a large retrospective analysis on 416 glioblastoma patients (Lacroix 2001). Age, performance status, extent of resection as well as contrast enhancement and necrosis on MRI were identified as independent prognostic factors. Patients who underwent a complete resection (> 98% as determined volumetrically on MRI) had a median survival of 13 months (95% CI, 11.4-14.6), compared to 8.8 months (95% CI, 7.4-10.2) for patients who had an incomplete resection. Nevertheless, despite attempts at radical resection, tumour recurrences occur almost invariably within the region of the previous resection cavity. A truly complete resection for this diffusely infiltrating disease is not possible.

Radiation therapy has been the backbone of glioma therapy for the last 25 years. The role of radiotherapy has been clearly established. Two trials by the Brain Tumor Study Group (BTSG) reported in 1978 and 1980 demonstrated improved survival with radiation therapy for malignant glioma (both grade III and grade IV tumors) (Walker 1978; Walker 1980). Although these trials could be criticized by today's standards (e.g. one fourth of the patients ineligible), they clearly showed improved survival in all arms incorporating adjuvant radiotherapy. Radiation therapy improved median survival from 4 to 8 months (Walker 1978), and increased the number of patients alive at 18 months from 10% to 15-20% (Walker 1980). Of note is that at that time whole brain irradiation was considered standard. In 1983 the Brain Tumor Cooperative Study Group (BTCSG trial 8001) began to randomize patients between WBRT of 60.2 Gy, and WBRT of 43 Gy plus a coned down tumour volume with 17.2 Gy. This trial showed no difference between the two radiation techniques (Deutsch 1989).
In a randomized trial by the Medical Research Council (MRC) focal radiotherapy of 60 Gy delivered in 30 fractions was superior to 45 Gy delivered in 20 fractions (Bleehen 1991). Dose escalation with a 10 Gy boost to the tumour volume did not increase survival in a randomized trial (Chang 1983). For patients with an adequate performance status a radiotherapy dose of 60 Gy in 30 daily fractions remains the standard of care on a type 1 level of evidence (Bleehen 1991).

The role of adjuvant chemotherapy has been investigated in a number of randomized trials since the mid 1970's. None of the prospective randomized trials to date has been able to demonstrate an unequivocal survival benefit. Although adjuvant BCNU (carmustine) is standard practice in the United States, there was only one prospective trial suggesting some additional benefit with BCNU ( Walker 1980). Nevertheless, a meta-analysis based on individual patient data showed a small, but significant improvement in survival for chemotherapy (Stewart 2002)
. Adjuvant chemotherapy with either a nitrosourea (BCNU) or procarbazine or high-dose prednisolone or BCNU and high-dose prednisolone were compared in a phase III trial. Overall there was no significant difference between the four treatment arms, however, the fraction of patients alive at 18 months was significantly lower for the corticosteroid only arm (15% compared to 23-29% for patients receiving chemotherapy (Green 1983).
One of the few positive trials was reported by the EORTC. Two hundred and sixty-nine patients were randomized to focal RT versus the same RT + dibromodulcitrol and BCNU (Hildebrand 1994). The longer survival in the experimental arm was mainly due to the subgroup of 29 (!) non-glioblastoma patients. Thus this study warranted confirmation, and in a subsequent randomized trial only patients with grade III glioma were eligible. This trial completed accrual but no outcome data are yet available.
In 1990 Levin et al. reported that adjuvant combination chemotherapy with procarbazine, CCNU and vincristine (PCV) provides superior survival to BCNU alone, based on a re-analysis of a randomized trial conducted between 1977-1983. The difference was only statistically significant for the subgroup of 73 anaplastic astrocytoma patients (total patients randomized 148) with a good performance status and who had received at least 1 cycle of therapy (Levin 1990). A retrospective pooled analysis of four RTOG protocols with a total 432 patients, however, failed to confirm a benefit of the PCV regimen over BCNU chemotherapy (Prados 1999). The MRC conducted a prospective randomized trial comparing standard focal radiotherapy with RT and up to 12 cycles of adjuvant PCV-chemotherapy (MRCBTWP 2001). Although this trial has been criticized for allowing different RT-schedules and for giving a modified PCV-regimen, it demonstrated conclusively the absence of a benefit for adjuvant chemotherapy in grade IV tumours but also in the subgroup of grade III anaplastic astrocytomas. ( MRCBTWP 2001).
Recently, Levin and colleagues reported their results of a randomized trial comparing adjuvant PCV versus adjuvant PCV and an oritine decarboxylase inhibitor (DFMO). Patients were only randomized after the end of radiotherapy and patients who had already progressed or with a decreasing performance status were thus not included in this trial. In patients with grade 3 anaplastic glioma median survival and time to progression were significantly longer in the PCV + DFMO arm (Levin 2003). Median survival times were 76 months and 61 months in the PCV/DFMO and PCV alone arms respectively.
Neoadjuvant administration of chemotherapy to patients with pure and mixed anaplastic astrocytomas was investigated in a randomized RTOG trial. Two hundred and ninety-one patients were randomized to either PCV chemotherapy followed by RT versus RT alone (Cairncross 2006). Although chemotherapy seemed to slightly prolong progression-free survival (2.6 v 1.9 years, p=0.53), it did not prolong overall median survival (4.8 and 4.5 years, respectively).
Based on the randomized data available, chemotherapy failed to consistently improve the outcome of patients with anaplastic astrocytoma. Outside a clinical trial, radiotherapy remains the sole standard adjuvant therapy for most patients on a type 1 level of evidence.

Chemotherapy is frequently prescribed in this situation, although no randomized data comparing cytotoxic therapy with best supportive care are available. it is, therefore, suitable for individual clinical use on a type 3 level of evidence. Most available agents have been tested against malignant brain tumours. No significant activity was demonstrated for paclitaxel (Fetell 1997; Chang 1998) and gemcitabine (Weller 2001), while some activity was suggested for camptothecins (Friedman 1999a; Friedman 1999b; Buckner 2003 ). Many chemotherapy agents are metabolized in the liver via the cytochrome P450 system. Unfortunately most of the classical antiepileptic drugs (carbamazepine, phenytoin, phenobarbital) are strong inducers of these enzymes, making a predictable dosing of the chemotherapy impossible or requiring several-fold increased doses. Another obstacle to chemotherapy for brain tumours is the blood-brain barrier. Thus sufficient drug concentrations may not be reached at the site of the infiltrative, non-contrast enhancing tumour. Camptothecins, and topoisomerase-1 inhibitors readily cross the blood-brain barrier and are good candidates for treatment of malignant glioma in patients not receiving enzyme-inducing antiepileptic drugs (EIAED). In patients with recurrent glioma objective responses have been reported for irinotecan (Campto?, Camptosar?) in 15 % of patients. In a phase I trial it has been demonstrated that in combination, both temozolomide and irinotecan can be escalated to their respective single agent maximally tolerated doses (Stupp 2003b). Over the last decade temozolomide has been specifically developed for patients with recurrent glioma (Stupp 2001). Three pivotal phase II trials were conducted for the development of temozolomide in recurrent glioma (Yung 1999; Yung 2000; Brada 2001). In contrast to most other reports on recurrent glioma, these trials were conducted separately for patients whose initial diagnosis was considered either glioblastoma multiforme (WHO grade IV) or anaplastic astrocytoma (WHO grade III). For these trials the, as yet unvalidated, endpoint of a progression-free survival at 6 months (PFS6mo) has been used and compared to a historical database (Wong 1999). To date there are no phase III data available demonstrating the superiority of temozolomide over other cytotoxic drugs. One randomized trial was designed as a non-comparative phase II trial and as such does not provide evidence for superiority of temozolomide ( Yung 1999). Nevertheless, it indicates a more favorable toxicity profile of temozolomide compared with procabazine, another alkylating agent. In patients with anaplastic astrocytoma and mixed oligoastrocytoma who recurred after prior RT and also prior chemotherapy, an overall objective response rate of 35% has been reported in 60% of the patients (Yung 1999). Responses were independent of prior exposure to other chemotherapy. The median progression-free survival was 5.4 months, the median survival 13.6 months. Survival was longer for the subgroup of responding patients. Due to the ease of administration and generally good tolerance, temozolomide is increasingly being prescribed for the treatment of brain tumours. Improved quality of life has also been shown for patients treated with temozolomide (Osoba 2000a). Additional benefit may be derived from closer and more continuous follow-up of patients receiving chemotherapy. All too frequently, patients have been prescribed high doses of corticosteroids at the initial diagnosis, which are then continued for an unnecessaryly long period of time. Subsequent neurological deterioration due to severe myopathy or secondary diabetes can be misinterpreted as tumour progression.

The recent introduction of small molecules targeting deregulated signalling pathways involved in malignant progression has opened the possibility of developing treatment protocols tailored to the individual molecular profiles of tumours. Inhibition of the PDGF-receptor and c-Kit with imatinib (STI571, Gleevec?) may be of promise. Furthermore, there are a number of targeted anti-angiogenic strategies in preclinical and clinical evaluation using small molecules such as VEGFR tyrosine kinase inhibitors (Yung 2003) or engineered antibodies directed against VEGF. It is today premature to conclude whether any of these agents either alone or in combination with conventional chemotherapy will be able to influence the outcome of patients with malignant glioma.

Based on the frequent local recurrence pattern and the protective effect of the blood-brain barrier, alternate routes of admistration of cytotoxic agents has been investigated. Biodegradable chemotherapy-impregnated polymers (wafers) have been developed, which are implanted in the surgically created resection cavity. The continuously released chemotherapy agent will then diffuse into the brain parenchyma. BCNU-wafers (Gliadel?) implanted in a selected group of patients with recurrent high-grade glioma (14% anaplastic astrocytoma) qualifying for a second resection was compared with placebo in a randomized trial and it is suitable for individual clinical use on type 2 level of evidence. There was a modest improvement in median survival of 7.2 months compared with 5.4 months ( Brem 1995). Placement of BCNU-wafers at the time of initial surgery failed to increase survival in a clinically meaningful way; it is, therefore, suitable for individual clinical use on a type 2 level of evidence (Westphal 2003). Two hundred and forty patients (9% anaplastic astrocytoma) undergoing surgery for malignant glioma were randomized to either BCNU-wafers or placebo, followed by postoperative radiotherapy. Median survival was 14 and 12 months, respectively (stratified log-rank p<0.03). Of note, was that the difference was only seen after 10 months and that there appears not to be a plateau of long-term survivors. Considering that all patients underwent at least debulking surgery, the outcome of both groups is disappointing. Ongoing trials now explore whether survival can be further improved by using polymers with a higher BCNU concentration (Olivi 2003). Drug delivery by intratumoral injection is limited by diffusion and increased parenchymal pressure. Direct positive pressure infusion allows for better distribution into the parenchyma through an interstitial convection flow. Some preliminary encouraging reports have led to large scale trials investigating this technique for a variety of agents including pseudomonas exotoxins linked to specific ligands (Laske 1997; Ram 2003; Stauder 2003).

Age has been identified as a major prognostic factor in malignant glioma. Primary glioblastoma is the most frequent brain tumour in this population. Due to the dismal prognosis, shorter hypofractionated radiation schedules haven been suggested (10 x 3 Gy, or similar). Ongoing randomized trials are comparing hypofractionated RT versus best supportive care (French trial) or hypofractionated RT v standard 60 Gy RT versus primary chemotherapy with temozolomide (Nordic trial). At the current time there are no randomized data available in this specific subgroup of patients who are mostly excluded from the larger randomized trials. In our phase II trial of newly diagnosed GBM patients the subgroup of patients of age 60-71 years were compared with patients between 50-60 years of age (Stupp 2003a ). No difference in outcome or treatment tolerance was noted. The performance status and neurological function are the most important factors to consider when deciding on therapy for elderly patients.

Cognitive and focal neurological deficits may have a great impact on long term survivors of brain tumors, regardless of the histology and grade of the tumors. Memory loss, apathy, concentration difficulties and personality changes may have a profound effect even in those patients who appear to have a Karnofsky performance status of 100. Surgery in the so called silent areas may contribute to cognitive deficits. Less clear are the late effects of radiation therapy on cognitive function. Radiotherapy is known to cause an early somnolence syndrome but may also cause late sequelae, in particular a delayed leuko-encephalopathy with cognitive dysfunction and radiation necrosis (Corn 1994; Crossen 1994; Kumar 2000. In individual patients it is difficult however to entangle the direct effects of the tumor on cognition from late effects of treatment. A recent survey on cognitive deficits in progression free survivors of low grade glioma failed to confirm the generally assumed relation between radiotherapy and cognitive deficits (Klein 2002). Only in those patients who had been treated with fraction of more than 2 Gy evidence of increased cognitive dysfunction was observed. The only other association with cognitive deficits was treatment with anti-epileptic drugs. Prior studies have suggested that whole brain radiotherapy may be associated with more cognitive deficits than involved field irradiation, but today involved field radiotherapy is standard practice (Gregor 1996). Radiation therapy may also affect cranial nerves, or induce endocrine dysfunction even in case of tumors distant from the hypothalamus-pituary region (Brandes 2000). Seizures may have a great impact on the quality of life even in patients with well controlled tumors. Newer anti-epileptic drugs may have less side-effects and should be considered, especially in those patients that are on a multi-drug regimen. Apart cognitive deficits a risk of death of 2.5% at 2 years has been reported for doses of 50.4 Gy. A risk of radionecrosis up to 5% in 5 years may occur after 60 Gy to one third or 50 Gy to two thirds of the brain volume or with 50-53 Gy to brain stem. Similar risk for blindness with 50 Gy to the optic chiasm. Also chemotherapy may induce late sequelae such as lymphoma or leukemia or solid tumors, lung fibrosis, infertility, renal failure, and neurotoxicity.

No general guidelines for the follow-up of OD can be given, these should be tailored to the individual patient taking tumor grade, previous treatments and remaining treatment options into account. Low grade glioma patients should be followed, even if the lesion is stable for many years. At some point in time, progression will occur and treatment should be installed before irreversible deficits occur.