Outcomes of presumed malignant glioma treated without pathological confirmation: a retrospective, single-center analysis

Abstract

Background

Tissue diagnosis is essential in the usual management of high-grade glioma. In rare circumstances, due to patient preference, performance status, comorbidities, or tumor location, biopsy is not feasible. Sometimes a biopsy is nondiagnostic. Many neuro-oncology clinics have patients like this, but these patients’ outcomes and responses to treatment are not known.

Methods

We retrospectively reviewed records from adult patients diagnosed with presumed high-grade glioma of the brain without definitive pathology, diagnosed between 2004 and 2016. We recorded several clinical variables including date of first diagnostic imaging and date of death.

Results

We identified 61 patients and subclassified them to brainstem glioma (n = 32), supratentorial presumed glioblastoma (n = 24), presumed thalamic diffuse midline glioma (n = 2), gliomatosis cerebri (n = 2), and cerebellar glioma (n = 1). Most brainstem glioma patients had no biopsy because of tumor location. Supratentorial presumed glioblastoma patients had no biopsy predominantly because of comorbidities. Median survival, from first diagnostic imaging, was 3.2 months (95% CI: 2.9 to 6.3 months) in the supratentorial glioblastoma group and 18.5 months (95% CI: 13.0 to 44.1 months) in the brainstem group. Treatment with radiation or chemotherapy did not alter the median survival of the supratentorial glioblastoma group (hazard ratio 1.41, uncorrected P = .5).

Conclusions

Patients with imaging diagnoses of high-grade glioma have similar, if not worse, survival than those with pathological confirmation. Based on these uncontrolled data, it is unclear how effective radiation or chemotherapy is in this population.

Standard-of-care management of high-grade glioma includes surgery for diagnostic confirmation because there is a differential diagnosis of brain lesions suspected to be cancer. Brain cancer may have actionable mutations. For these reasons, brain biopsy is recommended in cases of nonresectable suspected malignant glioma.¹ In select circumstances, tumor tissue is not readily available. Sometimes biopsy is too invasive: Patients with frailty, serious comorbidities, or brainstem lesions may decide not to have a brain biopsy. There are other cases for which a biopsy is performed but there is still no definitive diagnosis.

At our center (Princess Margaret Cancer Centre) we sometimes follow or treat patients who have suspected malignant glioma, but no tissue diagnosis. This practice of treating presumed brain tumors without tissue diagnosis used to be more common in the pre-MRI era. Most of what we know about the therapy of unbiopsied malignant glioma comes from then.^2–5 Some reports span both the pre-MRI and post-MRI eras.^5,6 Modern reports are confined to brainstem gliomas.^7–11 Because of advances in neurosurgical techniques, MRI imaging, and chemotherapy options, this field is due for an update in the modern era.

We suspect that patients without tissue diagnosis live shorter on average than those with tissue diagnosis, largely because of neurosurgical selection bias. We do not know how patients without tissue diagnosis respond to radiation or chemotherapy for their presumed malignant brain tumors in the modern era.

Methods

Inclusion Criteria and Data Collection

In a retrospective analysis that was approved by the institutional review board (19–5985) at University Health Network, we selected patients from the Princess Margaret Cancer Registry. Adult (age ≥ 18 years) patients diagnosed 2004 to 2016 with presumed glioma of the brain and no definitive pathology were included. Any patient who was ever seen at Princess Margaret Cancer Centre, either as an inpatient or an outpatient, is automatically entered into the Princess Margaret Cancer Registry. Patients who were seen at other institutions only would not be included in our database. The year 2004 was selected because this is when we started using temozolomide in Toronto; the year 2016 was selected to allow for mature survival outcomes. We then used neuroimaging to exclude patients with obvious low-grade tumors, for example, tectal gliomas and cortical nonenhancing tumors. Some brainstem glioma patients might have in retrospect been low grade but were not excluded because many nonenhancing brainstem tumors behave aggressively. Neither presumed gliomas of the optic nerve and chiasm nor tumors of the pituitary and pineal glands were included in the study. Patients with neuroimaging most consistent with lymphoma were excluded. Patients could have had a biopsy, but the resultant pathology must have been nondiagnostic.

Clinical characteristics that were collected on eligible patients included treatment with surgery, radiation, or chemotherapy. Using the clinical notes, Eastern Cooperative Oncology Group (ECOG) Performance Status was estimated at first visit. We extracted from the clinical notes reasons why no tissue diagnosis was attempted. Dates of symptom onset, first diagnostic scan date, date of death, and demographic data were collected from the clinical notes. The dates of death, where not otherwise available, were collected from public obituaries or patient relatives.

Statistical Analysis

Descriptive statistics were performed on the available demographic characteristics of patients with glioma without tissue diagnosis. Overall survival was defined as the time from first diagnostic imaging scan to the time of death. Survival data were censored at time of last contact. Statistical analyses were performed using R.¹²

Results

There were 1703 patients in the Princess Margaret Cancer Registry with a diagnosis of definite or suspected glioblastoma from 2004 to 2016. Of these, 76 patients fit our initial inclusion criteria. Of these patients, 15 were excluded: Seven actually had definitive pathology, 2 had imaging suspicious for lymphoma, 3 had nonenhancing cerebral lesions suspicious for low-grade glioma, 2 had tectal gliomas (and therefore were not thought to be high-grade lesions), and 1 had spine disease only. This left 61 patients (Figure 1).

Fig. 1.

Study Workflow

We then separated these 61 patients into 5 groups: (1) 24 patients with supratentorial, enhancing, suspected glioblastoma, (2) 32 patients with brainstem glioma, with or without enhancement, (3) 2 patients with thalamic tumors suspicious for diffuse midline glioma, (4) 2 patients with gliomatosis cerebri patterns of tumor, and (5) 1 elderly patient with a cerebellar enhancing tumor (Figure 1).

The average age at the time of diagnosis of patients was 73 years (range, 52-92 years) in the supratentorial group and 49 years (range, 18-82 years) in the brainstem group (Table 1). The brainstem enhancing subgroup was on average 57 years (range, 21-82 years) and the nonenhancing subgroup 40 years (range, 18-72 years). There were more men (38 patients) than women (23 patients) overall. The median ECOG performance status was 2 (range, 0-4) overall.

Table 1.

Baseline Characteristics, Treatments, and Survival of Patients

Characteristic	All patientsa(N = 61)	Supratentorial (N = 24)	Brainstem (N = 32)
Sex, No. (%)
Male	38 (62.3)	16 (66.7)	18 (56.3)
Female	23 (37.7)	8 (33.3)	14 (43.7)
Age at time of diagnosis, yb	61 ± 20	73 ± 12	49 ± 20
ECOG performance status, No. (%)c
0	1 (1.6)	0 (0)	1 (3.3)
1	21 (34.4)	5 (20.8)	14 (43.8)
2	17 (27.9)	8 (33.3)	8 (25)
3	14 (22.9)	8 (33.3)	6 (18.8)
4	8 (6.6)	3 (12.5)	3 (9.8)
Biopsy, No. (%)
Yes	7 (11.5)	1 (4.2)	5 (15.6)
No	54 (88.5)	23 (95.8)	27 (84.4)
Treatment, No. (%)
Chemotherapy + radiotherapy	21(34.4)	2 (8.3)	18 (56.3)
Concurrent only	4	1	3
Concurrent then adjuvant chemo	3	1	1
Radiation then adjuvant chemo	14	0	14
Radiotherapy only	33 (54)	16 (66.7)	14 (43.8)
Chemotherapy only	2 (3.8)	0	0
None	5 (8.2)	6 (25)	0
Time to treatment, wkd	3.3 (0.4-527)	3.5 (1-10.6)	3.4 (0.4-527)
Median survival, mo (95% CI)e	7 (4-15)	3 (3-6)	18 (13-44)
One-year survival, % (95% CI)e	41 (30-56)	0	66 (51-84)
Two-year survival, % (95% CI)e	28 (19-43)	0	46 (32-68)
Five-year survival, % (95% CI)e	15 (8-28)	0	22 (11-44)

^aEncompasses brainstem and supratentorial subgroups plus gliomatosis cerebri, cerebellar, and thalamic subgroups.

^bPlus–minus values are means ± SD.

^cEastern Cooperative Oncology Group (ECOG) performance status score at time of initial clinical assessment; scores range from 0 to 5 with higher scores indicating greater disability.

^dMedian values with range in parentheses.

^eSurvival defined as time from confirmatory diagnostic imaging to death.

Seven patients underwent brain tumor biopsy. None of the patients had surgical resection. None underwent autopsy. One of these biopsied patients had a brainstem lesion, 1 had a gliomatosis cerebri imaging pattern, and the other 5 were in the presumed supratentorial glioblastoma category. All of these pathology results were nondiagnostic (eg, gliosis not sufficient for diagnosis). For patients who did not undergo biopsy, we looked to the clinical notes to see why no biopsy was performed. In the supratentorial glioblastoma group, 37% had comorbidities that made biopsy less desirable, 37% had no biopsy because the imaging was classic, 37% because the patient was elderly, 26% because of strong patient preference, 21% because of poor performance status, and 16% because of tumor location. In the brainstem subgroup, in all cases the tumor location was a factor in the decision for no biopsy; other factors in up to 13% of cases included poor performance status, comorbidities, classic imaging, old age, and patient preference.

We had access to neuroimaging for 43 of the patients (Figure 2) and neuroimaging reports for the remainder. In the supratentorial glioblastoma group, all tumors had contrast enhancement. In the brainstem group, 17 had enhancement and 15 did not. In the thalamic diffuse midline glioma group, one had enhancement and one did not. In the gliomatosis group, one had enhancement and one did not. In the cerebellar group, the lesion was enhancing.

Fig. 2.

Representative MRI Images of Patients in Presumed Brainstem Glioma Group (Left), Presumed Supratentorial Glioblastoma Group (Middle), Presumed Thalamic High-Grade Glioma Group (Top Right), and Presumed Gliomatosis Cerebri Group (Bottom Right).

Only one patient is still alive (fourth column, third row, brainstem group).

In the supratentorial glioblastoma group, 18 patients received radiation and 6 did not. Two patients received temozolomide. Radiation treatments included intensity-modulated radiation therapy in 11 patients (most commonly 40 Gy in 15 fractions) and whole-brain radiotherapy in 7 patients (most commonly 20 Gy in 5 fractions). In the brainstem group, all patients received radiation therapy, 28 focal radiotherapy (most commonly 54 Gy in 30 fractions), and 4 whole-brain radiotherapy (most commonly 20 Gy in 5 fractions). Chemotherapy was given to 18 of 32 brainstem glioma patients: All of these patients received temozolomide (concurrent in 4 patients and adjuvant alone in the remainder). Second- and third-line chemotherapeutic agents were lomustine and etoposide, respectively.

One patient is still alive (Figure 2). He developed symptoms from his enhancing pontine lesion in 2008 and was treated with radiation. The lesion disappeared and has not recurred since. Dates of death are missing for 5 patients despite our attempts to contact family.

In the supratentorial glioblastoma group, the median ECOG performance status was 2. Those who received radiation or chemotherapy had a median ECOG of 2, and those who did not had a median ECOG of 3. In the brainstem group, the median ECOG performance status was also 2 and all patients received radiation. In the brainstem nonenhancing subgroup, the median ECOG was 1. Those who received chemotherapy had a median ECOG of 1.5 and those who did not, an ECOG of 2. Looking at all groups, those patients who received whole-brain radiation therapy had a median ECOG of 3 and an average age of 75 years (95% CI: 70-81 years). Those who received focal radiotherapy had a median ECOG of 2 and an average age of 54 years (95% CI: 48-61 years).

Survival curves for the supratentorial and brainstem group are depicted in Figures 3 and 4, respectively. Median survival from first diagnostic imaging was 3.2 months (95% CI: 2.9-6.3 months) in the supratentorial glioblastoma group and 18.5 months (95% CI: 13.0-44.1 months) in the brainstem group. Treatment with radiation or chemotherapy did not alter the median survival of the supratentorial glioblastoma group (hazard ratio 1.41, uncorrected P = .5). Median survival from start of radiotherapy was 2.4 months (95% CI: 1.9-5.8 months) in the supratentorial group and 17 months (95% CI: 10.8-32.8 months) in the brainstem group.

Fig. 3.

Overall Survival (From Time of Diagnostic Imaging to Death) in Presumed Supratentorial Glioblastoma Group.

Error margins are due to censored data for patients lost to follow-up.

Fig. 4.

Overall Survival (From Time of Diagnostic Imaging to Death) in Presumed Brainstem Glioma Group.

Error margins are due to censored data for patients lost to follow-up.

In the brainstem group, chemotherapy was not associated with prolonged survival, and the hazard ratio was 0.9 months (95% CI: 0.4-2.2 months). These survival curves are showed in Supplemental Figure 1. In the enhancing brainstem subgroup, chemotherapy was given to 7 of 17 patients (3 upfront and 4 on progression only). Survival outcomes for this subgroup can be seen in Supplemental Figure 2, in which there is a nonsignificant effect of chemotherapy.

Tumor enhancement was associated with shorter survival in the brainstem glioma group: 31.9 months without enhancement and 10.9 months with enhancement, as shown in Supplemental Figure 3. The patient in the gliomatosis cerebri group without enhancement lived much longer than the patient with enhancement.

Discussion

Our study shows that patients with presumed supratentorial glioblastoma who are followed in clinic without diagnostic confirmation tend to die sooner than those with confirmed glioblastoma. The Central Brain Tumor Registry of the United States 2016 data show a rate of 38% for 1-year survival for glioblastoma.¹³ Our study found no patients in the supratentorial subgroup alive at 1 year and a 41% rate for 1-year survival in all patients, indicating that our patients without tissue diagnosis are sicker than average patients. In randomized, controlled trials of radiation with or without temozolomide, the median survival of the control groups in adult and in elderly patients was 12.1 months in adults and 7.6 months in the elderly.^14,15 Notably, those who had biopsy rather than resection had shortened survival. For high-grade brainstem glioma, the median survival in a retrospective analysis spanning decades was 7 months.⁶ This is actually shorter than our brainstem group’s median survival, likely because our cohort included some low-grade brainstem glioma patients. Although we tried to exclude brainstem low-grade glioma from our cohort (eg, with tectal gliomas excluded), it is hard to identify these patients based on imaging characteristics alone.

We cannot know precisely why many of these patients had no biopsy. Many of our unbiopsied patients had a poor performance status and were older compared to those patients in clinical trials with verified histology.¹⁴ Several of our patients were referred for consideration of therapy because of progressive decline of neurological status. In some cases, patients were referred to our center from centers where they are uncomfortable treating without tissue. Extent of resection may be an important predictor of survival in glioblastoma. In a study involving 500 newly diagnosed glioblastoma patients, maximal surgical cytoreduction improves survival when extent of resection is 78% or more, and survival is even better when there is gross total resection.¹⁶ In a large randomized trial of elderly glioblastoma patients, patients with biopsy alone died at 1.67 times the rate of patients who had resection.¹⁵ All of these factors contribute to the poor survival of our supratentorial presumed glioblastoma cohort. It is possible that even patients with nondiagnostic biopsies fare worse than those with diagnostic biopsies of glioblastoma. Prominent tumor necrosis is associated with worse prognosis and also associated with a risk of nondiagnostic biopsy.

The biopsy of brainstem tumors can be fraught with difficulty and can lead to neurologic deficits. In diffuse intrinsic pontine glioma, for example, there are large practice variations in the decision to biopsy. One survey found that 14% of the respondents (mostly pediatric oncologists) biopsy all of their patients, whereas 42% biopsy their patients infrequently.¹⁷ There is no consensus on the need for biopsy in the management of presumed brainstem glioma.

There is inherent diagnostic uncertainty in patients without tissue diagnosis. There is a risk of treating someone for presumed glioma when he or she does not in fact have this diagnosis. Demyelination, infection, or inflammatory conditions are potential mimics of brain cancer. In 1965, a group of 21 patients with unbiopsied brain tumors was followed. Twelve underwent autopsy and all were found to have brain tumors.² Treatment without tissue confirmation is a reasonable option in some circumstances: Even in the pre-MRI era, patients treated for presumed brain cancer did in fact have brain cancer at autopsy.

When patients choose not to undergo biopsy, or when they are too ill to undergo biopsy, the decision to treat with radiation or chemotherapy becomes a difficult one. The only way to know how to treat patients with glioblastoma and poor performance status is with a randomized trial of therapies in this patient population. Until that time, the decision to treat patients with poor performance status or significant comorbidities requires a careful, individualized discussion with patients and their families. At our center, we generally offer therapy to patients with poor performance status only if patients consider their current quality of life acceptable—we do not wish to prolong patients’ suffering.

Most of the literature on unbiopsied glioma focuses on brainstem glioma. In those patients, several negative prognostic factors were found: age older than 50 years⁶ or 40 years,¹⁸ duration of symptoms 90 days or less before diagnosis, KPS 70 or less, and presence of contrast enhancement and MRI necrosis.¹⁸ Dey and colleagues found in an unbiopsied cohort that radiation therapy is not clearly associated with prolonged survival, but correlation ought not to imply causation.⁶ All of our brainstem glioma patients received radiotherapy.

There have been only 2 studies to our knowledge that looked at supratentorial high-grade glioma treated without tissue. Sheline and colleagues described patients diagnosed between 1934 and 1959 with glioma without biopsy.² There were 17 patients in the supratentorial subgroup; the 5-year survival rate was 24% (4/17).² The 5-year survival rate is so high because nonglioblastoma patients were included in the cohort. Lee described 41 adults and children with unbiopsied thalamic or brainstem gliomas.⁴ There were 14 patients with supratentorial tumors. Those with supratentorial tumors had better survival than the infratentorial group, probably because of the poor survival of diffuse intrinsic pontine glioma patients. The 5e-year survival rate was 42% in the supratentorial group, indicating that there were likely low-grade gliomas in this cohort.⁴ Both of these studies were in the pre-MRI era and are therefore only partly relevant to modern clinical practice.

It is clear that this cohort of patients, especially the supratentorial group, received less chemotherapy and less radiation therapy than guideline recommendations.¹ Only one patient in the supratentorial group received guideline-recommended concurrent then adjuvant temozolomide. Radiation doses were also generally less. These patients who were too sick to have a biopsy were too sick to receive standard-of-care therapy, and therefore could not derive the antineoplastic benefits of standard-of-care therapy.

Our study has several limitations: It is a retrospective cohort study at one institution. Patients with initially presumed diagnoses of high-grade glioma but who were subsequently confirmed to have other diagnoses would be missed in our study. Only with a prospective observational study could these patients be captured. We cannot make useful inferences about the efficacy of radiation or chemotherapy because we offered the treatments only to those we thought might benefit from therapy. We are missing imaging data on 30% of our patients, so we had to rely on MRI reports for them. Five patients were lost to follow-up. Finally, we have no data on patients during the same epoch who were not biopsied and were treated palliatively with no radiation or chemotherapy, never making it to our cancer treatment center. Comparison of our group to these patients might be useful.

Conclusions

Despite guideline recommendations to always obtain tissue diagnosis before treating for malignant glioma, there are still patients who are best managed without surgery. The poor survival of these patients compared to biopsied historical controls indicates that they probably do have malignant gliomas after all.

We herein have highlighted the clinical features and outcomes of patients followed in an adult neuro-oncology clinic who had no tissue diagnosis. The supratentorial presumed glioblastoma subgroup had on average a poor performance status, poor response to radiation or chemotherapy, and poor prognosis. These patients were unable to receive conventional doses and schedules of chemoradiotherapy and therefore fared very poorly. The brainstem glioma cohort had a better overall survival than the supratentorial cohort, likely because some low-grade gliomas were included.

Whether owing to tumor location, patient frailty, or comorbidities, it sometimes makes sense to treat without tissue. Both radiation and chemotherapy can be used to treat patients who have adequate performance status and presumed malignant glioma without pathological confirmation. These treatment decisions require nuanced assessments of patient preferences, clinical factors, and available evidence. We hope that our data can inform patients and clinicians who make these difficult decisions.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Conflict of interest statement. None declared.