Low-grade gliomas in older patients: a retrospective analysis of prognostic factors
- First Online:
- Cite this article as:
- Pouratian, N., Mut, M., Jagannathan, J. et al. J Neurooncol (2008) 90: 341. doi:10.1007/s11060-008-9669-3
- 99 Views
Low-grade gliomas (LGG) are increasingly being diagnosed in older patients (≥60 years). The rising incidence is incompletely understood but demands an analysis of the natural history and prognostic factors to determine if there are differences compared to younger populations. We retrospectively review a consecutive series of 20 older patients who presented between 1991 and 2004 with LGG to characterize the presentation, management, outcomes, and prognostic factors in older patients with LGG. Median follow-up was 27.3 months. Thirty-five percent of patients harboured oligodendrogliomas. Patients presented in equal numbers with seizure, focal neurological deficit, and mental status change. Median overall survival (OS) was 27.3 months. Patients who survive beyond 2 years experience prolonged periods of progression-free survival. Younger age, seizure presentation, and radiation treatment were prognostic of OS on multivariate analysis. Similarities with previous reports of LGG suggest that age should not affect the management of LGG patients. Prospective studies of older patients with LGG are needed to further characterize the optimum management of these patients.
KeywordsLow-grade gliomaGliomaElderlyPrognosisRisk factors
Approximately 2,000–3,000 low-grade gliomas (LGG, including World Health Organization [WHO] grade II astrocytomas, oligodendrogliomas, and oligoastrocytomas) are diagnosed in the United States every year, with a median age of diagnosis of approximately 40 years [1–6]. Gliomas, however, are increasingly being diagnosed in older patients [7–10]. The rising incidence in older patients is incompletely understood but has been attributed to an increased willingness to evaluate and treat older patients [6, 9, 11]. The presentation, natural history, prognostic factors, and treatment of LGG in older patients may be different than in younger patients, especially in light of the fact that increasing age is consistently identified as a poor prognostic factor [1–3, 12–18]. It is unclear how increasing age interacts with other prognostic variables in older patients and whether our understanding of LGG (largely based on the study of young to middle-aged adults) can be generalized to older patients. Characterization of the natural history and prognostic factors for these tumors in older patients is therefore critical to provide proper advice.
We retrospectively review a consecutive series of 20 patients 60 years of age or older who presented over 14 years to the University of Virginia to characterize the presentation (clinical and radiographic), management, outcomes, and prognostic factors in older patients diagnosed with LGG. We hypothesized a priori that, as in younger patients, older patients can experience long periods of progression-free survival but that overall survival (OS) and progression free survival (PFS) would be shorter than that reported in studies of younger patients and that increasing age and astrocytic histology would be poor prognostic factors.
Materials and methods
After obtaining Human Investigation Committee (HIC) approval, we retrospectively reviewed a consecutive series of 20 patients 60 years of age or older with histopathologically confirmed LGG who were diagnosed over 14 years (1991–2004) to the University of Virginia. Patients were identified by querying the University of Virginia Neuro-oncology Database using pathological diagnosis and age as search criteria. Hospital and departmental medical records were reviewed to extract information regarding patient demographics, medical history, radiologic characteristics, treatment, and clinical course.
Focal neurological deficit
Mental status change
Extent of resection
Gross total resection
Data analysis and statistics
The primary endpoints studied were OS and disease-specific PFS. Survival was measured from the time of initial presentation, even if tissue diagnosis was deferred to a later time. Progression was defined as either neurologic deterioration or radiographic progression, including increased area of FLAIR signal abnormalities and new or increased contrast enhancement. Factors evaluated for prognostic significance included age, gender, comorbidities (based on the Charlson comorbidity index [CCI]), presenting symptoms, radiological characteristics (including maximum tumor diameter, location, number of lobes involved, whether the tumor crossed midline, supra- or infra-tentorial location, radiographic pattern consistent with gliomatosis, or evidence of contrast enhancement), histology, and treatments (±chemotherapy, ±fractionated radiotherapy [XRT], and extent of resection). CCI was selected as a measure of comorbidity because it is an accurate predictor of short-term morbidity and mortality in older patients [20–22]. CCI ascribes a composite score based on consideration of a comprehensive list of comorbidities, including, but not limited to, cerebrovascular, cardiovascular, pulmonary, and hepatorenal disease as well as diabetes mellitus, dementia, hemiplegia, and other solid organ and haematological malignancies [20–22]. Extent of resection was based on the surgeon’s intraoperative impression, as determined from the operative note.
Actuarial OS was estimated using the Kaplan-Meier method. Multivariate modelling was based on the technique described by Collett and used by Pignatti and colleagues to identify prognostic factors in adults with LGG [23, 24]. Briefly, all potential prognostic factors were initially screened for association with survival using a univariate Cox regression analysis with a low threshold for inclusion (P < 0.10, based on likelihood ratios). Factors identified on univariate screening were then entered into a multivariate regression model. Factors with the least impact on the model (based on likelihood ratios) were then removed in a stepwise fashion until removal of the remaining factors resulted in a significant change in the model (P < 0.05). Then, all factors that were not selected for inclusion during the initial univariate screening were evaluated to determine if including these factors would significantly improve the multivariate model. Finally, after incorporating these additional factors, each factor was again tested to determine if any could be removed without significantly detracting from the overall model (P < 0.05). Results for each significant prognostic factor are reported in terms of hazards ratio (HR) with 95% confidence intervals (CI). Only significant prognostic factors are presented in tables.
Demographics and follow-up
Median follow-up was 27.3 months (range: 0.2–71.6 months). Although follow-up is seemingly short for LGG studies, analysis at this time was justified based on median OS having been reached for the study population and 70% of patients having been deceased at the time of data analysis. Patient demographics, mode of presentation, and presenting radiographic characteristics are described in Table 1. Six patients with a confirmed or suspected oligodendroglial tumor component were evaluated for chromosome 1p and 19q deletion. None had evidence of deletion. The Ki-67 LI ranged between 2% and 30% (mean: 12.7%, standard deviation: 8.4%). p53 staining patterns were classified as absent in 14 patients, rare in 1 patient, low in 1 patient, moderate in 1 patient, and high in 1 patient; p53 staining was not evaluated in 2 patients. Radiographically, maximum tumor diameter ranged between 1.8 cm and 10 cm, with a median of 6.8 cm.
Due to the retrospective nature of this study, treatment was not uniform across the patients studied, varying based on patient and physician preference and recommendations. Initial treatment included combined XRT and chemotherapy (20%), surgical resection alone (15%), surgical resection with postoperative XRT (10%), XRT alone (10%), chemotherapy alone (5%), and observation (biopsy only without surgical resection, chemotherapy, or XRT, 30%). The potential impact of various broad categories of treatment (e.g., chemotherapy and XRT) are assessed and described below. Details of therapeutic interventions (i.e., type and dosing of chemotherapy or radiotherapy regimen) are not provided due to the multiplicity of treatments used and consequently insufficient power to investigate the significance of each unique regimen. The impact of extent of resection (i.e., biopsy vs. subtotal resection vs. gross total resection) however was evaluated. Extent of resection for the study population is summarized in Table 1.
Median OS was 27.3 months (95% CI: 0–66.1 months). Six, 12, 24, and 60-month OS were 75%, 70%, 55%, and 39%, respectively. Death was due to neurological progression in all 12 patients in whom a definitive cause of death could be identified.
Prognostic factors for Overall Survival (OS) in months
Tumor crossing midline
Not seen or confirmed
Progression free survival
Median PFS was 15.2 months (95% CI: 0–57.6 months). Six, 12, and 24 month PFS were 70%, 60%, and 50%, respectively. Of 14 patients with progression, 2 initially had radiographic progression and 12 initially had clinical progression. All patients with progression died during the follow-up period.
Prognostic factors for Progression Free Survival (PFS) in months
Tumor crossing midline
Histopathological tumor progression was documented in 3 patients. In one patient, progression to an anaplastic glioma (WHO grade III) was diagnosed 35 months after initial presentation. In another patient, progression to glioblastoma multiforme (WHO grade IV, GBM) was diagnosed based on repeat biopsy 71 months after initial presentation. Finally, in one patient, who initially presented at the age of 72 with a multilobar glioma with contrast enhancement, surgical debulking within 1 month of initial biopsy revealed GBM, likely representing an initial misdiagnosis rather than a true progression.
An increasing incidence of histologically-confirmed brain tumors has been noted in all age groups, especially in patients who are 60 years of age and above [7–10]. Within this older group of patients, the increased incidence has been most pronounced in the diagnosis of oligodendrogliomas [9, 10]. Although the reason for the increasing incidence in older patients is incompletely understood, this change is likely related at least in part to a historical ambivalence and hesitation towards treating older patients with gliomas and a recent increased willingness of the medical community to pursue a diagnosis and treatment in older patients [3, 6]. In the past, older patients with LGG who presented with headaches or new onset seizures were likely managed symptomatically and never definitively diagnosed with a LGG. Accordingly, the increased rate of glioma diagnosis in older patients has coincided with an increased use of CT and stereotactic biopsies in this patient population between 1986 and 1994 [9, 11]. Others have proposed the increased rate of diagnosis is attributable to improvements in diagnostic technology (e.g., neuroimaging and intraoperative techniques) . Regardless of cause, the increased rate of LGG diagnosis in older patients necessitates characterization of the natural history and prognostic factors. This is the first study to critically analyze such factors exclusively in a consecutive series of older patients.
The profile of LGG in older patients resembles that reported in other studies. For example, there is a male predominance (approximately 1.5:1 male-to-female ratio) [3, 6]. The radiographic characteristics of LGG in older patients also resemble those previously reported in the literature, with the majority of tumors being supratentorial (90%), monolobar (60%, most commonly affecting the frontal and temporal lobes), and lacking contrast enhancement (75%) [2, 3]. The histopathological distribution of LGG in older patients (35% oligodendrogliomas) also likely reflects the increasing proportion of oligodendrogliomas reported in numerous studies [3, 9, 10, 26]. The increased incidence of oligodendrogliomas is unexplained but may be attributable to a bias to diagnose these tumors because they are associated with an improved prognosis . Although the WHO provides strict diagnostic criteria for gliomas, the subjectivity of LGG classification is highlighted by Castillo and colleagues who reported a surprisingly low 77% concordance rate amongst neuropathologists in interpreting all gliomas, with even lower concordance rates for specific subgroups, including astrocytoma, oligodendrogliomas, and particularly, mixed gliomas .
Although the presenting symptoms for older patients with LGG are qualitatively the same as that reported in other series (seizure, neurological deficits, altered mental status, and headache), the frequency with which patients present with various symptoms differs in older patients. Seizure is generally recognized as the most common mode of presentation for LGG (approximately 70–80% of patients) . In older patients, however, seizure, focal neurological deficit, and mental status change each account for 25–30% of presenting symptoms. While older patients presenting with focal neurological deficits and altered mental status are most likely to have had a stroke, these results suggest that LGG (and other neoplasms) should remain in the differential diagnosis of older patients presenting with these symptoms.
According to the National Center for Health Statistics, between 1995 and 2004, median survival for a 65-year-old person ranged between 17.4 and 18.7 years and for a 75-year-old person between 11.0 years and 11.9 years . The median OS of older patients with a diagnosis of LGG (6.0 years in 60–69 year olds and 1.3 years in 70–79 year olds) is therefore certainly shorter than the general population. Although the median OS for older patients with LGG is longer than that reported for GBM in the elderly (~12 months), it is lower than that reported in other series of LGG (36–95 months) [1–3, 14, 30]. Decreased median OS in older patients is consistent with numerous studies that have reported that increased age is a robust negative prognostic factor [1–3, 12–18]. Likewise, the 5-year survival (37%) reported in this series is in the lower range of that reported in the literature for LGG (27–68%) [1–3, 14]. By studying age as a binomial (e.g., comparing patients less than or greater than 40 years of age) rather than as a continuous variable, previous studies have understated the negative impact of each additional year of age on prognosis, especially beyond 60 years of age. The study demonstrates a 23% increased death rate for every additional year of age beyond 60 years of age at the time of diagnosis. Pignatti and colleagues similarly found a linear relationship between age and prognosis, cautioning that the age threshold of 40 years that they used in their study should not be interpreted as an absolute cutoff .
The significant impact of age on outcomes is highlighted by comparing median survival in the current series of older patients with that described by Pignatti and colleagues . Based on one data set and validated on another, Pignatti and colleagues devised a 5-point prognostic scoring system that is a reliable predictor of median survival. They reported a median OS of 7.72 years for low-risk (0–2 points) and 3.20 years for high-risk (3–5 points) patients. In the current series, the median OS for low- and high-risk patients were 1.56 and 0.48 years, respectively, both considerably shorter than previously reported. The shorter survival of both risk groups suggests the prognostic scoring system described (based on an age threshold of 40 years) may be insufficient to predict survival in older patients. On the other hand, the shorter survival of high-risk patients compared to low-risk patients suggests that at least some of the same risk factors (tumor diameter, histology, tumor crossing midline, and the presence of neurologic deficit) continue to affect outcomes even in older patients (see discussion of prognostic factors below) .
Like younger patients with LGG, some older patients diagnosed with LGG experience prolonged periods of PFS and OS. Claus and Black reported that in patients over 64 years of age with LGG, they observed a rapid decline in survival in the first 2–3 years after diagnosis but that mortality leveled off in subsequent years, “suggesting a subgroup of patients has a more positive prognosis than the rest” . Our results confirm this observation. While 25% of the patients died within 6 months of diagnosis, the next quartile of patients survived until 27 months after diagnosis. Furthermore, the third quartile of patients experienced greater than 5-year survival (precise estimate unattainable since not all patients have died). Therefore, in older patients diagnosed with LGG, patients who survive beyond the first 6 months (especially those who survive beyond 24 months) can be expected to experience significant PFS and OS, similar to younger patients with LGG. The disparity in early and late death rates raises the possibility that the patients who experience short survival (e.g., less than 6 months) actually had higher grade tumors that were surgically under sampled or histologically misinterpreted. Stereotactic biopsies generally have reported accuracy rates of ~80%, but as low as 51% when interpreted by someone other than a dedicated neuropathologist [31–33]. Sampling error is in fact confirmed in one patient in this series, who underwent early craniotomy and resection after initial biopsy, revealing higher grade histology. This patient was not excluded from the current analysis (despite proven sampling error) because such patients are important to our understanding of the management of older patients diagnosed with LGG, even if they do not actually harbour LGG. The histopathological progression reported at 35 and 71 months, however, likely represent true progression, corresponding to the time when malignant transformation is normally seen .
LGG prognostic factors in older patients
Schiff and colleagues recently extensively reviewed prognostic factors for LGG . The most important negative prognostic factors to emerge from the literature include increasing age, astrocytic histology, large tumor diameter (greater than 5 or 6 cm), tumors crossing midline, and the presence of neurologic deficits [2, 3, 12, 14, 23]. In contrast, presentation with seizures (which largely occur in patients who are otherwise neurologically intact) is often identified as a positive prognostic factor [2, 13–15]. Seizure presentation, neurologic deficit, and performance status are inter-related; reports of each being related to outcome likely represent the same underlying phenomenon affecting survival. Other prognostic factors that are inconsistently identified include extent of tumor resection, gender, race, and treatment parameters [2, 3].
Even in this subset of older patients, increasing age was the most robust negative prognostic factor. Presentation with seizure and XRT were also significant prognostic factors on multivariate analysis. Although not significant on multivariate screening, univariate screening also identified tumor crossing midline, CCI, and astrocytic histology as negative prognostic factors. Unlike prior studies, tumor size and extent of resection was not significantly associated with prognosis on either univariate or multivariate analysis. We speculate factors reported as important in previous studies were not significant for one of two reasons. First, because of the small sample size, this series may have been underpowered to detect these effects on multivariate analysis. For example, post hoc exploratory analysis confirms that in this small series, patients presenting with seizures were significantly more likely to have unilateral disease (i.e., tumors not crossing midline) (P < 0.05). Likewise, patients presenting with seizures tended to have smaller tumor diameters than other patients (approached significance, P = 0.07). Because of the limited sample size, we could not independently evaluate the effect of each factor on prognosis. Alternatively, the other prognostic factors identified (i.e., increasing age) may be more important in older patients than in other groups of patients, minimizing the contribution or significance of other putative prognostic factors, as described in the earlier discussion of the overwhelming impact of age on prognosis. Likewise, CCI, which has been validated as a tool for predicting survival in older patients, may not correlate with survival because the overriding diagnosis of LGG overwhelmed the predictive ability of CCI.
Histology and molecular markers
Assessing the effect of histologic subtype on prognosis was limited by surgical sampling issues. Diagnoses were based on needle biopsies in 75% of the patients in this series resulting in a large fraction of patients (30%) with a diagnosis of LGG NOS and not appropriately assessable for the impact of histology on prognosis. Nonetheless, consistent with previous reports of improved prognosis in patients with oligodendrogliomas, tumors with confirmed astrocytic histology (astrocytomas and oligoastrocytomas) had significantly worse OS than others on univariate analysis [12, 23]. The failure to identify this effect on multivariate analysis is presumably due to limited power and sampling issues.
The diagnosis of all cases was based on strict criteria of the WHO classification . The current WHO criteria for dividing infiltrating gliomas into grade II and grade III rely primarily on the assessment of proliferative activity, by means of mitotic activity. Because of the limitations in histological classification, we undertook an analysis of the proliferative activity by measuring the Ki-67 LI and p53 expression. Ki-67 LI analysis was available in 75% of cases. Ki-67 LI was higher in this series (mean of 12%) than that reported in the literature for LGG (usually below 6%). Fifty-three percent and 93% of patients had Ki-67 LI greater than 5% and 10%, respectively, thresholds previously reported to be associated with poor prognoses [35, 36]. Ki-67 LI may not be of prognostic significance in this study because of the relatively higher Ki-67 LI seen in this patient population, making it a less reliable indicator of survival. Although such high Ki-67 LI may imply misdiagnosis, patients with some of the highest Ki-67 LI in our series (20.80%, 12.90%, and 9.40%) were amongst the longest survivors (41 months [still alive], 63 months [still alive], and 72 months [deceased]), suggesting these patients did not actually have higher grade tumors which would have been associated with much shorter OS. The discrepancy between the histological grading and the Ki-67 LI may be a reflection of the ubiquitous heterogeneity seen in infiltrating gliomas and heterogeneity-induced sampling errors that may limit diagnostic accuracy, particularly in small biopsies . The rate of p53 overexpression (10%), however, was lower than reported in younger patients with LGG (30–50%) [38, 39]. Nevertheless, like Ki-67 LI, p53 expression pattern was an unreliable predictor of survival. Although p53 mutation and overexpression have been associated with poor outcomes, the 2 patients with immunohistochemical patterns suggestive of p53 mutation survived 66 and 72 months [38, 40]. Loss of heterozygosity (LOH) of chromosomes 1p and 19q has also been implicated as a prognostic factor . Chromosomal deletions were not identified in any of the patients in this series making it impossible to evaluate the prognostic significance of this marker. However, the absence of chromosomal deletions in any of the 6 patients evaluated suggests that chromosomal deletions may be less common in older patients [42, 43]. The absence of deletions may contribute to overall worse outcome of older patients with LGG. The current results suggest that we may not be able to rely upon chromosomal deletion status to stratify older patients with LGG.
Due to the retrospective nature of this study, treatment was not uniform across the patients studied, making it difficult to ascertain the efficacy of various treatment modalities. Nonetheless, the use of XRT was associated with significantly improved OS. While this may be a true effect of treatment, this result is susceptible to selection bias. Twenty-five percent of patients died within 6 months of diagnosis, none of whom received XRT. However, their short survival is likely not attributable to their lack of treatment, but due to their clinical status at the time of presentation (e.g., intracerebral hemorrhage). Despite numerous randomized controlled trials evaluating the dose and timing of XRT, the benefit of XRT for LGG with respect to OS (compared to not receiving any radiation) is based on retrospective studies [12, 17, 44]. As there will never be a study that precludes patients from receiving XRT, it is and will be impossible to prove conclusively that XRT is actually beneficial for OS. Nevertheless, as it is a standard of care for patients with LGG, this study supports the continued use of this treatment modality in older patients with a diagnosis of LGG, keeping in mind that older patients may however be at greater risk of neurotoxicity from radiation than younger patients. Surprisingly, XRT did not provide a significant benefit with respect to PFS, as described by van den Bent and colleagues in the EORTC 22845 trial . The efficacy of other treatments, including surgical resection and chemotherapy in older patients is not supported by the current analysis. In fact, 4 of the 5 longest survivors had biopsy only, never having undergone a resection procedure. Keles and colleagues recently critically reviewed the literature with respect to the impact of extent of resection upon outcomes in LGG, finding that a preponderance of modern studies support extensive resection over biopsy alone . Therefore, further investigation into the impact of extent of resection on outcomes in older patients with LGG may be warranted.
The study is limited by small sample size and its retrospective nature. Characterization of a relatively rare entity, however, requires such an analysis with these limitations. As with all retrospective studies, treatment is not uniform, thereby complicating the analysis of outcomes. This study is also limited with respect to the assessment of the impact of extent of resection on outcomes. Due to retrospective nature of the study, extent of resection had to be gleaned from surgeon’s impression as detailed in operative notes. A more standardized approach to assessing extent of resection (based on comparison of pre- and post-operative MRI) would be helpful in determining the impact of surgical resection on PFS and OS. Nonetheless, the current analysis provides an initial analysis and identification of prognostic factors. In light of the increasing attention being given to the care of older patients, this type of analysis, in spite of the limitations, is necessary if we are going to begin to provide fair and accurate counseling to this patient population. Future studies should aim to analyze a larger number of patients, preferably in a prospective manner, grouping patients across multiple institutions to ensure adequate enrollment.
Although LGG has long been considered a disease of young adults, LGG are increasingly being diagnosed in older patients (≥60 years of age). The rising incidence is likely due to an increased willingness to investigate and treat disease in older patients. While older patients present with similar symptoms as younger patients, a greater proportion of older patients with LGG may present with altered mental status and focal neurological deficit. While median survival is 27 months, patients who survive 2 years beyond initial diagnosis may experience prolonged periods of progression-free survival. Despite being diagnosed with LGG, those who do not survive beyond the initial 6 months may actually harbour higher grade lesions that were misdiagnosed due to sampling error. As with the young adult population, increasing age continues to be prognostic of poor survival. Presenting with seizure and XRT likely portend a survival benefit. Similarities between this subgroup and previous reports of LGG suggest treatment strategies should not necessarily be modified solely based on the age of the patient. Prospective studies of older patients with LGG will be needed to further characterize the optimum management of these patients.