Introduction

Diffuse astrocytoma (DA) is one type of low-grade glioma (LGG) and classified as WHO (World Health Organization) grade II. DA is a relatively slow-growing brain tumor with a median survival time of 3.9 to 10.8 years.1 2 3 4 5 6 However, DA infiltrates brain parenchyma, making it hard for total resection. The rate of gross-total resection is only 14 to 17%.2 3 Therefore, residual tumors become the burden of patients with DA in the long-term follow-up.

Low-grade glioma (LGG) comprised of several types of tumor, that is, astrocytoma, oligodendroglioma, and mixed oligoastrocytoma. Associated factors for poor outcomes in patients with LGG are older age, Karnofsky’s performance status (KPS) < 70, prior neurological deficits, tumor size >6 cm, tumor crossing midline, and nonhemispheric/noncerebellar tumor.6 7 8 Postoperative radiotherapy (RT) is yet a debatable predicting factor for prognosis of LGG. In a study, postoperative conventional RT tends to shorten time to progression and prolonged progression-free survival but not the overall survival.9 Another study shows no significant difference in survival between pediatric patients with LGG who received postoperative RT and those who did not.8 Nonetheless, there has been limited data about prognosis among subpopulation of patients with LGG including patients with DA.

Nomogram has been used to predict clinical outcome in various groups of diseases, such as malignancies,10 11 degenerative diseases,12 and metabolic diseases.13 From the literature review, there is the heterogeneity of histology in the group of LGG and lack of evidence of the prognostic factors that focuses on diffuse astrocytoma. We aimed to identify prognostic factors associated with survival of patients with diffuse astrocytoma. This study developed the nomogram to predict survival of DA hoping to use in the clinical practice.

Subjects and Methods

Results

The clinical manifestations of the 64 patients with DA are shown in Table 1. More than half of the DA was slightly dominant in males. The mean age was 39.4 years (SD = 17.3). The patients usually presented with a seizure and a progressive headache. Additionally, 96.9% of patients had the Glasgow coma scale (GCS) score 15, while 3.1% had a GCS score of less than 15. The common tumor location involved the frontal and temporal lobe in 37.5 and 23.4%, respectively. The thalamic tumor was found in 1.6% of cases. Moreover, the mean tumor volume was 5.6 (+SD = 1.8) and 92.2% of the DA was solitary tumors while 7.8% was multiple tumors.

In this study, the total resection rate was 12.5%, while the rates of subtotal, partial resection, and biopsy were 10.9, 42.2, and 34.4%, respectively. Most of the patients (82.8%) underwent radiotherapy after resection.

Survival Analysis

The prognosis of the diffuse astrocytoma was pitiable as the Kaplan–Meier curve in Fig. 1A Mean follow-up time was 42 months (SD = 34.3). Also, the overall median survival time was 44 months (95% confidence interval [CI]: 31.0–87.0), while the 1-, 2-, and 5-year survival probability were 85.9, 67.6, and 42.3%, respectively. Moreover, the malignant transformation was observed in 17.2% of cases.

Close

Fig. 1 Survivals of patients with diffuse astrocytoma glioma by predictors. (A) The overall median survival time was 26 months (95% confidence interval [CI]: 19.0–41.0). (B) A comparison of survival times among age groups (log-rank test, p = 0.004). (C) A comparison of survival times among patients according to motor response groups (log-rank test, p < 0.001). (D) A comparison of survival times among biopsy and resection operations (log-rank test, p = 0.04).

Export to PPT

Fig. 1 Survivals of patients with diffuse astrocytoma glioma by predictors. (A) The overall median survival time was 26 months (95% confidence interval [CI]: 19.0–41.0). (B) A comparison of survival times among age groups (log-rank test, p = 0.004). (C) A comparison of survival times among patients according to motor response groups (log-rank test, p < 0.001). (D) A comparison of survival times among biopsy and resection operations (log-rank test, p = 0.04).

According to Cox’s proportional hazard regression analysis as shown in Table 2, the significant parameters for increased death were age 60 years or more group (hazard ratio [HR] = 4.83, p = 0.001), motor response of GCS score less than 6 in groups (HR 39.49, p = 0.009), positive hypervascular sign (HR = 2.05, p = 0.03), biopsy (HR = 0.47, p = 0.03) in univariate analysis. By multivariable analysis, the significant model consisted of age 60 years or more (HR = 5.76, p < 0.001), the motor response score of GCS less than 6 (HR = 75.47, p = 0.003), and biopsy (HR = 0.45, p = 0.02).

Table 2

Factors associated with the death of patients with diffuse astrocytoma

	Univariate analysis		Multivariable analysis
Factor	Odds ratio (95%CI)	p-Value	Odds ratio (95%CI)	p-Value
Abbreviations: CI, confidence interval; Ref., reference. aData show only “yes group” while reference groups (no group) are hidden. bEloquent area defined tumor involved motor cortex, sensory cortex, visual center, speech center, basal ganglion, hypothalamus, thalamus, brainstem, dentate nucleus.
Gender
Male	Ref.
Female	1.16 (0.61–2.20)	0.63
Age (y)
< 60	Ref.
≥ 60	4.83 (1.86–12.55)	0.001	5.76 (2.09–15.91)	< 0.001
Aphasiaa	4.50 (0.56–36.0)	0.15
Seizurea	1.08 (0.56–2.08)	0.80
Motor response of GCS
Equal 6	Ref.		Ref.
Less than 6	39.49 (2.47–631.49)	0.009	75.47 (4.15–1369.4)	0.003
Preoperative Karnofsky performance status
< 80	Ref.
≥ 80	1.29 (0.67–2.51)	0.43
Location
Frontal lobea	1.06 (0.55–2.05)	0.84
Temporal lobea	0.83 (0.49–1.68)	0.69
Parietal lobea	0.67 (0.20–2.28)	0.52
Brainstema	1.17(0.15–8.71)	0.87
Thalamus/basal gangliona	2.85 (0.65–12.32)	0.16
Corpus callosuma	2.20 (0.63–7.71)	0.21
Sellar/suprasellar regiona	0.20 (0.02–1.75)	0.14
Spinal corda	2.59 (0.32–20.52)	0.36
Lateralization of tumor
Left	Ref.
Right	1.45 (0.73–2.88)	0.28
Midline	1.36 (0.53–3.48)	0.51
Eloquent area a,b	0.87 (0.46–1.65)	0.68
Number of tumors
Single	Ref.
Multiple	2.18 (0.75–6.28)	0.14
Positive hypervascular signa	2.08 (1.04–4.12)	0.03	1.81 (0.88–3.75)	0.10
Leptomeningeal disseminationa	1.59 (0.37–6.73)	0.52
Preoperative hydrocephalusa	1.09 (0.51–2.30)	0.81
Midline shift
< 0.5	Ref.
≥ 0.5	1.20 (0.59–2.45)	0.59
Maximum diameter (cm)
< 3	Ref.
≥ 3	0.73 (0.30–1.78)	0.49
Type of operation
Biopsy	Ref.		Ref.
Resection	0.47 (0.23–0.95)	0.03	0.45 (0.21–0.92)	0.02
Postoperative Karnofsky’s performance status
< 80	Ref.
≥ 80	0.98 (0.52–1.85)	0.95
Radiotherapy
No	Ref.
Yes	0.86 (0.35–2.06)	0.73

As the Kaplan–Meier curve in Fig. 1B D, the median survival time of age less than the 60-year group was 34 months, whereas age 60 or above group was 7 months (log rank test, p < 0.01). The patients with motor response score of GCS less than 6 had median survival time at 40 months while motor response score of GCS equal to 6 had median survival time at a month (log-rank test, p = 0.005). Additionally, the median survival time of biopsy and tumor resection was 22 months and 27 months, respectively (log rank test, p = 0.04).

Nomogram Development and Performance Test

The nomogram was developed using the significant parameters in the multivariable analysis as shown in Fig. 2. Additionally, the calibrate plot revealed that the model was nearby the ideal and had a bias-corrected concordance-index value of 0.933. The nomogram is simple in general practice. For example, a 58-year-old patient (no point) had six scores in the motor response of GCS (no point), and the MRI of the brain showed a left frontal diffuse astrocytoma with positive of the hypervascular sign, and he underwent to total tumor resection (no point) as Fig. 3. Therefore, total points equal to 0 points which approximately corresponds to more than 90% of 1-year survival probability, more than 70% of 2-year survival probability, 35 to 40% of 5-year survival probability, and 80 to 90 months of predicted-survival time.

Close

Fig. 2 Nomogram predicting 1-year, 2-year, 5-year survival probabilities (Sur. Probs.) and median survival time (months). To use the nomogram, draw a straight line upward from the patient's characteristics of age group, motor response, type of surgery to the upper points scale, the sums of the scores of all variables. Then, draw another straight line down from the scale of the total points through the 1-year, 2-year, 5-year, and median survival times. This is the probability of the presence of prognosis in an individual.

Export to PPT

Fig. 2 Nomogram predicting 1-year, 2-year, 5-year survival probabilities (Sur. Probs.) and median survival time (months). To use the nomogram, draw a straight line upward from the patient's characteristics of age group, motor response, type of surgery to the upper points scale, the sums of the scores of all variables. Then, draw another straight line down from the scale of the total points through the 1-year, 2-year, 5-year, and median survival times. This is the probability of the presence of prognosis in an individual.

Close

Fig. 3 Brain MRI of a 58-year-old patient who survived at 4-year-follow-up. (A) Axial T1-weighted image shows left hypointense frontal mass. (B). Axial T2-weighted image demonstrates the intratumoral flow void sign. (C). Axial T1-weighted image shows vivid enhancement after contrast injection. MRI, magnetic resonance imaging.

Export to PPT

Fig. 3 Brain MRI of a 58-year-old patient who survived at 4-year-follow-up. (A) Axial T1-weighted image shows left hypointense frontal mass. (B). Axial T2-weighted image demonstrates the intratumoral flow void sign. (C). Axial T1-weighted image shows vivid enhancement after contrast injection. MRI, magnetic resonance imaging.

For validation, the performances of nomogram were evaluated as the binary classifiers instead of survival probability. Therefore, nomogram needed to find the optimal cut-off point for highest performance. At an optimal cut-off point of 18, the nomogram was validated by self-consistency test that sensitivity, specificity, PPV, NPV, accuracy, and AUC were 0.77, 0.61, 0.25, 0.94, 0.70, and 0.70, respectively for predicting 1-year mortality as Fig. 4. Nevertheless, nomogram performance gradually dropped in 2-year and 5-year prediction as Table 3.

Close

Fig. 4 Receiver operating characteristic (ROC) curve and area under the curve (AUC) of predicted mortality as binary classifiers using cut-off-point of 18. (A) 1-year predicted mortality, (B) 2-year predicted mortality, and (C) 5-year predicted mortality.

Export to PPT

Fig. 4 Receiver operating characteristic (ROC) curve and area under the curve (AUC) of predicted mortality as binary classifiers using cut-off-point of 18. (A) 1-year predicted mortality, (B) 2-year predicted mortality, and (C) 5-year predicted mortality.

Table 3

Performance test of a nomogram predicting 1-, 2-, and 5-year mortality of diffuse astrocytoma at cut-off 18 points

Prediction	Sensitivity	Specificity	PPV	NPV	Accuracy
Abbreviations: NPV, negative predictive value; PPV, positive predictive value.
1-year mortality	0.77	0.61	0.25	0.94	0.70
2-year mortality	0.60	0.63	0.42	0.77	0.62
5-year mortality	0.53	0.65	0.60	0.58	0.59

Discussion

Overall median survival time have been reported between 3.9 and 10.8 years in the LGG which composed of astrocytomas, oligodendrogliomas, and mixed oligoastrocytomas.1 2 3 4 5 6 Spych et al reported LGG had the median survival times of 26.9 months (range: 25.4–46.8 months), whereas the median survival times was 48 months (range: 30 months–138 months) according to the study by Kumthekar et al.2 3 Moreover, The 1-year, 3-year, and 5-year overall survivals of LGG were 88.3, 55.1, and 43.3%, respectively.2 As the present results, the prognosis of our cohort was poorer than prior studies that the median survival time of the cohort was 26 months and the 5-year probability was 22.0%. Since the heterogeneity of the LGG population limits to comparison with the present cohort.

For negative prognostic factors, Pignatti et al reported that age >40 years, whereas Okamoto et al reported age >50 years associated with the poor prognosis.6 Similarly, the present study found the elderly patients had a poor prognosis. Furthermore, the presence of neurologic deficit before surgery was one of the significant prognostic factors, while motor response score of GCS was less than 6, as well as the poor performance status of the patients was associated with worse outcome in the present study.

Several studied reported that tumor resection significantly impacts the survival of LGGs.23 24 25 The result in the present study concordances the prior studies that biopsy was significantly associated with increased mortality. Furthermore, the effect of RT is controversial for improving outcome and prognosis. LGGs treated with postoperative RT which was highly correlated with overall survival3 9 in prior studies, whereas other studies reported that postoperative RT was no statistically significant difference in survival was seen between the postoperative RT and non-RT groups.2 8 26 However, the heterogeneity of the study population of LGGs was observed in previous studies. In the present cohort of DA, the radiotherapy following surgery was no statistically significant association with overall survival rate. From the literature review, a few studies focused on the prognosis of the DA group since the limitation of sample size were considered. Nevertheless, the present study, which focused on the specific DA group, had the strength in the homogeneity of the study population. To the best of our knowledge, this is the first study mentioned to evaluate and validate nomogram which is specific to DA.

Gorlia et al proposed nomogram for predicting survival of patients with LGGs including astrocytoma, oligodendroglioma, and mixed oligoastrocytoma. The final overall survival model, independent prognostic factors were identified as time since first LGG symptoms (HR = 0.67, p = 0.009), Medical Research Council (MRC) neurological score (HR = 1.51, p = 0.0001), independent confirmation of astrocytoma (HR = 1.96, p < 0.001), and tumor size (HR = 1.74, p = 0.001). However, there is a limitation of validation because the prediction of nomogram is probability at each time point and survival time.27 For example, predicted the 1-year probability of nomogram is 50% what we should interpret these results for an individual in the real-world applications. Therefore, we proposed nomogram validation as binary classifiers each time point with the optimal cut-off point in the present study. The nomogram of our cohort had acceptable performances for predicting 1-year mortality that had a high level of sensitivity and accuracy. For general practice, the nomogram could be applied as the screening tool for advising patients and managing treatment strategies. However, nomogram’s performances dropped for predicting 2- and 5-year mortality that needs external validation in the future.

Limitations

Certain limitations of the present study should be acknowledged. The IDH1 mutation of DAs did not perform in the present study because these genetic investigations have not routinely estimated in Thailand. Moreover, the possibility of bias and confounding factors cannot be excluded from the retrospective study. However, we presented to adjust the model by multivariable analysis for controlling this limitation28

Conclusion

We provided nomogram predicting prognosis of a patient with DA. The nomogram was acceptable performance for predicting 1-year mortality. The tool is a good clinical utility for optimizing therapeutic approaches and counseling patients.