Supratentorial low grade astrocytoma: prognostic factors, dedifferentiation, and the issue of early versus late surgery
- Departments of Neurosurgery and Neuropathology, University Hospital Rotterdam and Departments of Neurology and Medical Statistics, Dr Daniel den Hoed Cancer Center, Rotterdam, The Netherlands
- Dr MLC van Veelen, Department of Neurosurgery, Academic Hospital Rotterdam, Dr Molewaterplein 40, 3015 GD Rotterdam, The Netherlands. Telephone 0031 10 463 9222; Fax 0031 10 463 3735.
- Received 12 September 1996
- Revised 15 August 1997
- Accepted 20 November 1997
BACKGROUND A retrospective study of patients with low grade astrocytoma was carried out because the best management of such patients remains controversial. Prognostic factors were identified by multivariate analysis. Special attention was paid to the effect of extent and timing of surgery.
METHODS Ninety patients with low grade astrocytoma were studied. Seventy two patients had resective surgery, 15 had a diagnostic biopsy only, and three patients had resective surgery after initial biopsy.
RESULTS Significant prognostic factors for survival were age, preoperative neurological condition, epilepsy as the single sign, extent of surgery, and histology. The extent of surgery was highly significant on univariate analysis (p=0.002); however, after correction for age and preoperative symptoms this was considerably reduced (p=0.04). A subgroup of 30 patients with epilepsy as their single presenting symptom was identified. Thirteen of these patients were treated immediately after diagnosis, whereas the other 17 patients were initially followed up and treated only after clinical or radiological progression. Survival in both groups was identical (63% survival rate after five years) and much better than survival for the whole group (27% survival rate after five years). Malignant dedifferentiation was observed in 25 (70%) of 36 patients who were reoperated, after a median period of 37 months. This period was 41 months for the subgroup of patients with epilepsy only and 28 months for the remaining patients.
CONCLUSIONS Due to the retrospective nature of the study only restricted conclusions can be drawn. Low grade glioma with epilepsy as the single symptom has a much better prognosis than if accompanied by other symptoms. This prognosis is not influenced by the timing of surgery. It seems, therefore, safe to defer surgery until clinical or radiological progression in low grade glioma with epilepsy only.
The best management of patients with low grade astrocytoma remains controversial.1-3 The timing of surgery, the extent of surgery, and the need for radiotherapy continue to be a matter of discussion despite many retrospective studies.
The timing of surgery became an issue after the introduction of CT. Due to its better sensitivity, CT disclosed lesions in an earlier stage than scintigraphy. Early diagnosis by CT raised the question as to whether early operation in patients with no or minor neurological deficit would benefit the duration and quality of survival, considering the rather benign course of the disease and the risk of postoperative neurological damage. The fact that in most patients extensive surgery is a favourable prognostic factor4-10 is often used as an argument in favour of early surgery. However, the two retrospective studies that specifically considered the issue of deferral of therapy did not show clear differences in survival nor in the rate of malignant transformation between early or late surgery.11 12 We analysed prognostic factors in 90 patients with low grade astrocytoma. Treatment strategy changed over time due to the increasing availability of CT, introduction of stereotactic biopsy, and fluctuation in opinion on the preferred management. These changes resulted in two treatment strategies for the 30 patients with epilepsy as their single presenting sign. One group was treated early and one group was initially followed up and treated after clinical or radiological progression. We studied the factors that influenced the choice and outcome of either strategy.
Materials and methods
This study comprised all 90 patients with a histologically confirmed low grade astrocytoma who had their first operation or biopsy between 1975 and 1989. Patients who were less than 18 years of age at operation were excluded. Only supratentorial tumours were included. No patient was lost to follow up.
Age, sex, presenting symptoms, preoperative duration of symptoms, location, and CT characteristics of the lesion, preoperative and postoperative performance status, type, and extent of surgery, as well as complications, histology, radiotherapy, and survival were taken from the patients’ charts.
The performance status was scored before operation and at discharge from the hospital using a neurological function classification on a four point scale13:
0=able to work, with only slight or no neurological deficit (includes mild headache, epileptic insults controlled with anticonvulsants, minimal aphasia, minimal hemianopia, slight hemiparesis, sensory disturbances); 1=able to be at home, with minor neurological deficit (includes headache, epileptic insults, some apathy permitted, mild aphasia, hemianopia, hemiparesis grade 4 (MRC scale)); 2=requiring hospital admission, with major neurological findings (includes organic brain syndrome, severe headache, aphasia, hemiparesis grade 2–3, sphincter disturbances, wheelchair dependent); 3=requiring hospital admission and in serious physical or neurological state (includes diminished consciousness or coma, severe aphasia, hemiplegia (0–1), severe ataxia, bedridden more than 50% of the time).
All charts contained a fairly extensive report on the operation, which allowed us to estimate the extent of tumour resection using the following scale: 1=stereotactic biopsy (performed since 1983); 2=partial resection (estimated tumour removal<75%); 3=gross total removal (estimated tumour removal⩾75%).
Postoperative CT to more reliably estimate the extent of tumour removal was not available on a routine basis.
All histological slides were reviewed by the neuropathologist. Low grade astrocytoma according to World Health Organisation (WHO; 1993) criteria as well as low grade oligodendrogliomas and mixed types were included. Pilocytic astrocytomas and low grade gliomas in the posterior fossa were excluded.
Available options for therapy were early surgery, biopsy only, or deferral of surgery, with or without initial diagnostic biopsy. Deferral of surgery or the “wait and see” policy was considered as a feasible option only for patients with epilepsy as the single presenting symptom. In the case of a “wait and see” policy, surgery was performed after clinical or radiological progression. In an attempt to identify a potential bias the arguments for or against either therapy choice were recorded from the patient’s charts. In most cases the reasoning as to which treatment was chosen was not extensively described. Therefore we first defined possible arguments and then scored them as “present” or “not present”, further adding arguments as we encountered them in the charts.
In general, patients were reoperated on if recurrence occurred more than one year after initial surgery. In case of dedifferentiation of tumour, radiotherapy was applied. Radiotherapy was also administered in patients with an inoperable recurrence.
The follow up period of this study was censored at 31 December 1993. At that time the patient’s condition was recorded as either alive, dead, or as hospital mortality (death within 30 days after surgery, from surgical or medical complications). Date of recurrence was scored at the time of increase of symptoms or of radiological progression of the tumour after surgery.
Survival curves and mortality rates were calculated according to the actuarial method of Kaplan and Meier. Greenwood’s formula was applied to calculate the standard error (SE) of the survival rate estimates. In addition to the overall survival duration we calculated the progression free survival, which was defined as the survival until the first date of recurrence or death. For the survival analysis the overall survival duration and the progression free survival were measured from the day of first surgery. However, when comparing the survival from immediate and deferred surgery in the subgroup of patients with only epilepsy, survival duration was measured from the day of the first positive scan, because this was considered, retrospectively, as the day that the decision was made between immediate or deferred surgery.
The Cox proportional hazards model was used to determine the patient and tumour characteristics that were most related to prognosis, measured by the duration of survival and the progression free survival. All factors mentioned in table 3 as well as contrast enhancement on CT and duration of symptoms were fed into the Cox univariate analysis. The associated log likelihood ratio test was applied to identify differences or trends in survival and progression free survival between groups, in multivariate analysis.
Ninety patients met all inclusion criteria. Table 1 lists the patients’ characteristics preoperatively. Mean age was 41 (SD 12) years. Sixty one patients (68%) presented with epilepsy and in 30 of them epilepsy was the only symptom. Most tumours were located in one or more cerebral lobes. One was located in the thalamus and there was one ventricular tumour. Six lesions were bilateral. Table 1 gives the CT characteristics. Interestingly, a considerable number of CT lesions (34%) showed contrast enhancement. Seven patients were initially diagnosed by scintigraphy.
Of the 30 (33%) patients with epilepsy as the single presenting symptom 13 were selected for immediate surgery, whereas for the remaining 17 patients a “wait and see” policy was chosen. Eventually, all patients in the “wait and see” group underwent surgery. The median interval between the first positive scan and surgery was 14 months (mean 29 months), with a maximum of 138 months. The median interval for the group with immediate surgery was one month (mean two months).
The remaining 60 patients who presented with neurological deficit (with or without epilepsy) were operated on without actual deferral, one month (median) after the first positive CT.
At the time of surgery one third of the patients had a good (grade 0), one third an intermediate (grade 1), and one third a poor performance status (grade 2 or 3). The performance status improved after operation in 27 patients: 28% of the patients with preoperative performance status grade 1 and 63% of the patients with preoperative performance status grade 2 or 3. Four patients had a lower performance status after operation.
The surgical procedure was restricted to a biopsy in 18 patients. Three of them underwent decompressive surgery later (after nine, 21, and 70 months). Seventy two patients had a more or less extensive tumour resection. Only in 13 of these 72 patients an estimated resection of more than 75% was achieved. Nine patients had medical complications and three had surgical complications. Four patients died within one month of surgery.
Thirty six patients were reoperated on at least once. Fifteen of these patients had a third operation and four a fourth operation. In another six patients the symptomatic regrowth was primarily (mono)cystic and they only had a drain implanted in the cyst.
Thirty one patients received radiotherapy at recurrence either because the recurrent tumour was deemed inoperable or because histology at reoperation showed dedifferentiation to a higher grade of malignancy.
In most patients histology disclosed a low grade astrocytoma (table 2). (None showed mitotic activity, cell polymorphism, or necrosis, only moderate cell density was accepted.) Oligodendroglioma and mixed types of tumours occurred less often. Thirty six patients were reoperated at least once. In 25 of them a higher grade of malignancy was diagnosed at either the second, third, or fourth operation. This occurred after a median period of 37 months (actuarial estimate; range 3–84 months) for the whole group, 41 months for the patients with epilepsy only, and 28 months for the other patients. Within the group of patients with epilepsy only this period was 40 and 43 months, respectively for the deferred and immediate treatment group. In 11 of the 36 reoperated patients no dedifferentiation was noted. Their reoperation was after a median period of 19 months.
SURVIVAL AND PROGRESSION AFTER SURGERY
Of the 90 patients 82 showed progression during follow up and 75 patients died (table 3). The overall survival probability for the whole group was 27% (SE 5) at five years and 14% (SE 4) after 10 years. The median survival time was 36 months.
Table 3 shows the five year overall survival and progression free survival probabilities for subgroups with different features; figure 1shows the corresponding survival curves. The patterns in overall survival and progression free survival were similar. Favourable characteristics were younger age (fig 1A), epilepsy as single presenting symptom (fig 1B), and oligodendroglioma (compared with astrocytoma). The absence of contrast enhancement was not identified as a favourable characteristic. The duration of preoperative symptoms did not show a significant trend with overall survival or progression free survival over the whole follow up, but fig 1C shows that the death rate in the first years was much higher in the group of patients with a preoperative duration of symptoms<three months. This analysis, however, should primarily be regarded as exploratory because most of the subgroups are small and some of the characteristics are interrelated.
Multivariate Cox regression analysis identified age over 50 years (relative death rate (RDR)=2.5, p=0.002), absence of epilepsy (RDR=3.5, p=0.001), and three or more symptoms before surgery (RDR=2.2, p=0.007) as the most important factors. Given these factors patients with oligodendrogliomas showed a lower death rate than the patients with astrocytomas. However, the difference was not significant (RDR=0.5, p=0.06). Grade of malignancy, preoperative performance status, nor location of the lesion showed a significant association with survival, given age, and symptoms.
The patients whose tumours could be decompressed extensively fared much better than the patients who had just a biopsy or a small decompression of their tumour. However, the differences in survival related to extent of surgery seem to be partly evidence of selection. The 13 extensive resections were performed primarily in younger patients (12 patients<50 years) presenting with epilepsy (eight patients with epilepsy only, three patients with epilepsy and one other symptom) who had a right sided tumour (10 of 13). Also, a large resection occurred more often among patients with an oligodendroglioma (five of 14) than among patients with an astrocytoma (seven of 72). Without adjustment for age and symptoms the difference in survival between patients with a large resection and the rest was highly significant (p=0.002), but this was reduced considerably after adjustment for age and symptoms (p=0.04).
Radiotherapy was not evaluated as a prognostic factor, because the indications for radiotherapy were variable, and would probably produce a highly biased value.
“WAIT AND SEE” VERSUS IMMEDIATE SURGERY
The patients who were operated on immediately were compared with those with a “wait and see” policy. Figure 2 shows the overall survival, calculated from the date of the first positive CT or scintigraphy, for both treatment options in the subgroup epilepsy only (the overall survival for the other patients with neurological deficit is shown as well). There is no clear difference between both treatment groups; they have a pooled five year overall survival of 63 (SE 9)% and a 10 year overall survival of 32 (SE 9)%. The estimate of the relative death rate in the immediate surgery group compared with the “wait and see” group was 1.02 (95% confidence interval 0.4–2.5). This result has to be interpreted cautiously. We tried to identify potential biases, firstly, in the arguments for either therapy found in the charts and secondly, in different patient characteristics of both treatment groups. Arguments to choose the “wait and see” option were (more than one argument for each patient possible): reluctance of the patient to surgery (one) no neurological deficit (15), diffuse and extensive hypodensity on CT (10), and lesion located in eloquent areas (seven). Reasons to operate after initial delay were the patient’s wish (one), incontrollable seizures (two), neurological deficit (10), progression on CT (seven), and acute tumour haemorrhage with signs of herniation (one). Arguments for early surgery were the patient’s wish (one), the need for histological verification (six), contrast enhancement on CT (three), supposed arachnoid cyst on CT (one), and location favourable to surgery (10). Clearly, the reasons for deciding on either treament strategy are numerous and diverse. Most important arguments are firstly the operability of the lesion—that is, location of the lesion and diffuse growth pattern, and secondly the expected malignancy. These arguments are partially reflected in the differences in the patients’ characteristics (table 4). Early treated patients were younger (mean age 33 v 41 years), had more right sided lesions (12 of 13 v eight of 17), and were operated on more extensively. Contrast enhancement on CT occurred slightly more often in the early treated group (four of 13 versus three of 17). Thus the early treated group tended to have the most favourable patient characteristics for survival. Therefore, the lack of a better survival rate in the early treated group seems not to be produced by a preponderance of unfavourable patient characteristics in this group.
Survival rates and favourable prognostic factors in this study are comparable with series reported previously4-12 14-21(table 5), although the five year survival rate is lower than in most studies. This could be explained by the relatively high age in this group and the high percentage of small tumour resections.
Practically all studies identify age as a strong independent prognostic factor. Patients under the age of 20 years have a five year survival rate of about 80%, between age 20–40 the survival rate decreases to about 50%, whereas survival rates for patients above 40 years of age are around 30%.
Extent of surgery is another favourable prognostic factor in most studies. It should be realised, however, that extent of surgery may be subject to inaccuracy and bias. The surgeon’s estimation of the size of resection is known to be of limited accuracy. Postoperative CT was not done routinely in most series. Deep seated lesions, older patients, or patients with gross neurological deficit possibly tend to be operated on less extensively. Furthermore, in several studies,4 8-10 15-18 operative deaths were not included in their analysis. However, for assessment of the true effects of surgery it is more realistic to include the associated mortality. Reported percentages of death related to surgery in the 1970s and 80s vary between 2% and 9%. Finally, extensive surgery seems to be especially effective in younger patients,4 16 but this finding may have been influenced by the higher number of pilocytic astrocytomas in the younger group.16 In considering the effect of extensive surgery, the findings of Vertosick et al 19 are surprising. His patients received radiotherapy after verification of histology by stereotactic biopsy only. He reported a median survival of 8.2 years in a group of 25 patients. These results, however, can at least partially be explained by the young age of 35 years of his population and the very low percentage of enhancing lesions on CT.
About half of the reports indicate that radiotherapy is a significant prognostic factor. Laws et al,4Garcia et al,15 and Shaw et al 16 found that radiotherapy benefits primarily older patients. Several studies show a favourable effect only for subtotally removed tumours.4 15 20 22 Medbery et al 7 and Shaw et al 16established a relation between dose of radiation and duration of survival. However, a recent prospective trial failed to find a dose response to 45 versus 59 Gry.21
In our study the initial symptom of epilepsy only is a stronger prognostic factor than age. Although the preoperative neurological condition is a significant prognostic factor in most studies, epilepsy only was either not identified as a prognostic factor or not evaluated separately. Only Laws et al 4 and Northet al 5 found epilepsy as a favourable prognostic factor on univariate analysis. However, for anaplastic gliomas, epilepsy has been recognised as a favourable prognostic sign in numerous studies.12 23-26 In our opinion patients with epilepsy only should be considered as a separate group within the low grade glioma. They have a better prognosis, which, in combination with younger age and intact neurological condition, may well have implications on choosing the best strategy.27 The dilemma of early versus late treatment particularly applies to this group. Early surgery entails a certain risk of morbidity and mortality which may be acceptable if it would produce a substantially better survival. Without surgery these patients seem to have, in most cases, several years of good survival quality with epilepsy as their only symptom.
Two retrospective studies11 12 considered the management of patients with epilepsy and unverified lesions on CT. Early treatment was compared with deferred treatment. Differences in outcome could not be identified, neither in survival nor in the time until malignant dedifferentiation. It is difficult to explain these findings considering that Laws et al 4 and many others reported that extensive surgery produces a better survival. However, it may be erroneous to extrapolate extensive surgery to early surgery.
In our study, early versus late treatment in patients with epilepsy only did not show a clear difference in survival calculated from the time of the first CT. However, our results should be interpreted with some caution. Both groups comprise few patients. Patient characteristics in both groups were not evenly distributed, although the equal survival rate for both groups cannot be explained by the presence of favourable patient characteristics in the group with deferred treatment. Finally, patients were selected for this study based on histology and not on CT. Patients with a “wait and see” policy who were later, on initial surgery, found to carry more malignant lesions, are not included in this study. Excluding these patients from the analysis may have favoured the outcome of the “wait and see” group. Although we are aware of the presence of potential bias, we think that our study shows that deferring treatment is safe for patients with “epilepsy only”.
Because of the few patients with “epilepsy only” we did not try to identify subgroups, based on additional prognostic factors such as age, for which deferral of treatment would be less or perhaps more appropriate. Based on the known prognostic factors it may be may argued that for the group of patients over 40 years of age a more aggressive approach may be safer. Their prognosis is worse, and although they probably do not benefit as much from extensive surgery they benefit more from radiotherapy than the younger patients. Moreover, about half of the non-enhancing lesions in patients over 45 years of age seemed to have a malignant lesion grade III or IV.28 Therefore, tumour verification seems to be warranted in this group either by resective surgery or biopsy, followed by radiotherapy. The younger patients, on the other hand, have a better prognosis, the need for radiotherapy is not clear, and non-enhancing lesions are highly associated with low grade gliomas. Tumour verification would not alter treatment strategy. In these patients a “wait and see” policy is in our opinion fully justified until better figures come available.27 It is necessary, however, to be aware of the risks of deferred therapy such as tumour haemorrhage, rapid and unexpected deterioration, and also incorrectly labelling the patient as inoperable. Resective surgery should be reconsidered in every patient showing clinical or radiological progression.
Definition of the best policy for low grade astrocytoma should be based on prospective studies. Our study shows that the outcomes after early or late surgery in patients with epilepsy alone are not clearly different and a prospective study comparing both these options seems warranted. Based on our figures we calculated that the minimum size of the study group totals 250 patients to prove a difference in survival of 20% and 800 patients to prove a difference of 10%. Considering the incidence of low grade glioma and its survival, only a multicentre trial may answer these questions.