Incidental diffuse low-grade gliomas: A systematic review and meta-analysis of treatment results with correction of lead-time and length-time biases

Abstract

Background

Better overall survival (OS) reported in patients with incidental diffuse low-grade glioma (iLGG) in comparison to symptomatic LGG (sLGG) may be overestimated by lead-time and length-time.

Methods

We performed a systematic review and meta-analysis of studies on adult hemispheric iLGGs according to the PRISMA statement to adjust for biases in their outcomes. Survival data were extracted from Kaplan–Meier curves. Lead-time was estimated by 2 methods: Pooled data of time to become symptomatic (LTs) and time calculated from the tumor growth model (LTg).

Results

We selected articles from PubMed, Ovid Medline, and Scopus since 2000. Five compared OS between patients with iLGG (n = 287) and sLGG (n = 3117). The pooled hazard ratio (pHR) for OS of iLGG to sLGG was 0.40 (95% confidence interval [CI] {0.27–0.61}). The estimated mean LTs and LTg were 3.76 years (n = 50) and 4.16–6.12 years, respectively. The corrected pHRs were 0.64 (95% CI [0.51–0.81]) by LTs and 0.70 (95% CI [0.56–0.88]) by LTg. In patients with total removal, the advantage of OS in iLGG was lost after the correction of lead-time. Patients with iLGG were more likely to be female pooled odds ratio (pOR) 1.60 (95% CI [1.25–2.04]) and have oligodendrogliomas (pOR 1.59 [95% CI {1.05–2.39}]). Correction of the length-time bias, which increased the pHR by 0.01 to 0.03, preserved the statistically significant difference in OS.

Conclusions

The reported outcome in iLGG was biased by lead-time and length-time. Although iLGG had a longer OS after correction of biases, the difference was less than previously reported.

Glioma is a primary neoplasm that had been classified according to its morphological similarities with putative cells of origin in the brain. The World Health Organization (WHO) classification system for glioma is shifting to a molecular diagnosis using IDH mutation (IDHm) and chromosome 1p/19q codeletion (Codel), both of which represent the main diagnostic and prognostic markers. In the new classification released in 2021,¹ which classifies IDHm astrocytoma as grade 2 to 4, tumors that were previously defined as IDW-wild (IDHw) astrocytoma are classified in glioblastoma if they meet the following three criteria: The concurrent gain of whole chromosome 7 and loss of whole chromosome 10, TERT promoter mutations, or EGFR amplification.

Diffuse low-grade gliomas (LGGs) are slowly growing tumors that usually cause seizures as an initial symptom without neurological deficits. Although they are a low-grade malignancy, the majority are expected to cause malignant progression.^2,3 Although the 10-year malignant progression-free survival (MPFS) rate was 38.3% in patients with IDHm tumors without Codel and 62.5% in those with Codel, MPFS further declined with time.³ The initial tumor volume and degree of resection were significantly associated with progression. Therefore, early maximal safe resection preserving the eloquent brain areas is currently considered a better treatment option.^2,4

Although the prevalence of incidentally found LGG (iLGG) is very low (0.064% of imaging studies),⁵ the frequency is increasing with the spread of neuro-imaging utilizing MRI.⁶ Treatment decisions in iLGG are often difficult because of the uncertainty of the diagnosis, a lack of high-level evidence to support the treatment, possible complications of treatment, and patient anxiety. Few authors reported the treatment results of iLGG. Whereas all reported that overall survival (OS) in a series of patients who received early treatment for iLGG was better than that in symptomatic LGG (sLGG),^6–11 the lack of prospective randomized studies meant that no definite conclusions could be drawn. In general, the early detection of cancer by screening improves patient survival. However, the interpretation of treatment results is often biased.^12,13 One source of bias is the overestimation of survival time due to earlier detection by screening in comparison to the clinical presentation (lead-time bias). Since gliomas are not a target of screening, the actual lead-time for iLGG is the interval between the incidental diagnosis and the occurrence of symptoms, the lead-time to symptoms (LTs). Although the bias was pointed out in a previous systematic review of the treatment outcomes of iLGG,¹⁴ corrections of the bias have been never attempted. Moreover, no randomized studies compared survival between early and delayed surgeries in iLGG. Only one retrospective study reported the surgical timing, (surgery performed before or after the occurrence of symptoms), which had no significant effect on OS or MPFS in iLGG.¹⁵ For a valid comparison of the effect of early treatment in iLGGs we have to subtract the lead time from the overall survival time in patients with iLGG. Disappointingly, the LTs have rarely been reported. It is necessary to accumulate the LTs data to estimate the lead time. Pallude et al. reported that the median LTs was 48 months (mean 55 months) in 13 patients with iLGG.⁷ Another method for correction of the lead-time bias is applying biological tumor growth models that calculate the necessary time to grow to become symptomatic for iLGG.^16,17 For this analysis, it is necessary to determine the speed of iLGG growth.

Another predisposition is the length-time bias (length-biased sampling); neoplasms that grow more slowly are more likely to be detected incidentally.¹³ Although some authors reported the dominance of oligodendrogliomas in iLGG,^9,11 it is still controversial.^5,9 On the other hand, it is usually difficult to correct a length-time bias. Duffy et al. developed a two-tumor type model to correct the length-time bias in cancer screening; one tumor type being a rapidly growing tumor and another being significantly slower, with each having different probabilities of being detected by screening or symptomatic ones.¹² If a difference in the frequency of tumor type exists between iLGG and sLGG, it would be applicable to correct the bias.

The aims of this study were to reveal biases that affect the interpretation of treatment results in previous studies of iLGG and to find more reliable results. Because each of the previously reported studies analyzed relatively small numbers of patients, a systematic review and meta-analysis would be an appropriate method to find reliable treatment results in relation to iLGG. One problem was that targeted tumors were defined by older WHO classification systems (majority by WHO 2007 or 2016). Therefore, where possible, we collected molecular data.

Methods

The study was conducted according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement and registered in PROSPERO (CRD42021293390).

Literature Search and Data Extraction

The PRISMA search flow diagram is outlined in Figure 1. We searched for relevant English articles using the keywords “low-grade glioma,” “astrocytoma,” or “oligodendroglioma” and “incidental” OR “asymptomatic” OR “natural history” in PubMed, Ovid Medline, and Scopus from 2000. The literature search of each database and data extraction were conducted independently by 2 of the authors. Studies with supratentorial adult iLGG were included. Pediatric cases, familial cancer disease, and studies limited to a specific location were excluded. In the case of articles from the same institute, the newer article or articles with the necessary data were selected for data extraction. Although we excluded studies that had clinical data of less than 10 cases of iLGG, even case reports were included to accumulate LTs data. Incidental tumors should not have neurological symptoms related to the mass or seizures. OS and progression-free survival (PFS) was separately recorded from a diagnosis or from initial treatment. The diagnostic method for malignant transformation was recorded as clinical (radiological) or histological. Cases with a diagnosis of higher-grade glioma were considered as malignant transformation if they initially showed typical imaging features of low-grade glioma and were observed for ≥2 years without symptoms. The included studies were finally determined by discussion with the authors.

From each study, we collected data on patient age, sex, tumor size, growth rate, location (eloquent area or not), histology (including molecular data), and outcome (OS, PFS, and MPFS). The LTs and observation period (time from diagnosis to treatment) were recorded when available. When only the median value and range were available, the mean and standard deviation were calculated using the method of Hozo et al.¹⁸ The pure oligodendroglioma rate was calculated as the number of oligodendrogliomas divided by the total number of cases without oligoastrocytomas. Treatment-related factors, including the rate of total removal, and the number of patients with adjuvant radiotherapy and chemotherapy were also recorded. Neurological complications were recorded as transient, permanent, and epilepsy (early or late).

We used the hazard ratio (HR) to determine the effect of treatment in the meta-analysis. When the HR was not available, we calculated the missing values according to the method of Tierney et al.¹⁹ or from the data extracted from the Kaplan–Meier curves. When there was no death, we entered “0.01” as assumed death to calculate the HR. We used WebPlotDigitizer (https://automeris.io/WebPlotDigitizer/) to extract data from Kaplan–Meier curves. We verified the accuracy of the extracted data by overlapping a curve from the extracted data on an original curve in the PowerPoint software program. The data extraction from the curve were performed by one author and accuracy was checked by another. Adjustment was repeated until a complete overlap was confirmed. In one study,⁷ the Kaplan–Meier curve, which showed 5 deaths, was not concordant with the description of the number of deaths (4 deaths at a median of 8.9 years; range, 2.3–13.5 years; mean, 8.4 years). In this case, we reconstructed the curve according to the description.

Correction of Lead-Time and Length-Time Bias

For the estimation of lead time, we employed 2 methods. In the first method, we collected the LTs data and the pooled mean data were used as the lead time. In the second method, we calculated the time necessary for an incidental tumor to become symptomatic. Because the majority of studies employed the velocity of diameter expansion (VDE) as the growth rate and the ellipsoid method for the calculation of tumor volume, the tumor diameter was approximated as follows.

$Diameter=2×volume3$ (reference²⁰)

ΔD, the difference of the mean diameter between preoperative incidental and symptomatic tumors; mVDE, the mean VDE obtained from the pooled data.

The estimated lead time was subtracted from the survival time by the method described by Duffy et al. (Supplementary File 1).¹²

k, the mean time for incidentally diagnosed tumors to be symptomatic (pooled mean LTs) or calculated time in a tumor growth model (LTg).

t, observed survival time or time to last follow-up.

Correction of the length-time bias was performed by the 2 tumor type model described by Duffy et al.¹² The assumption was that the difference in probability of death from 2 tumor types during the observation period was related to different probability in screening and symptomatic detection in the 2 tumor types. We first investigated the difference in the frequency of pure oligodendroglioma between iLGG and sLGG. Then, we calculated the true relative risk of death from the tumor in question for incidentally detected tumors versus symptomatic tumors, independent of the length bias. The precise methodology was described in Supplementary File 1.

Statistical Analysis

We used the R software program (v4.03) (https://www.r-project.org/) to perform the statistical analyses (the packages of Meta, Metafor, and EZR). An inverse-variance approach with a random-effects method was applied for the meta-analysis. The reviewed studies were tested for heterogeneity (I²statistic). The metamean function was used to synthesize the mean VDE and LTs values. A single-arm meta-analysis was performed to synthesize the iLGG data using a metaprop or metarate function. A generalized linear mixed model was used to synthesize proportions or incidence rates in the presence of 0 event counts. The results of the odds ratio (OR) for the difference between iLGG and sLGG and the HR for the time-to-event outcome difference in each factor were pooled. Meta-regression analyses were performed to identify factors related to heterogeneity. The analyses were performed using the rma.uni function in the Meta software package. In Kaplan–Meier curves from reconstructed data, the log-rank test, and Cox-proportional hazard test were applied for the survival analysis. Two-sided P values of <.05 were considered to indicate statistical significance.

Risk of Bias

We used the JBI critical appraisal checklist for the case-control study (Supplementary File 2) (https://jbi.global/critical-appraisal-tools) to evaluate the risk of bias in 5 studies that compared OS between iLGG and sLGG; this evaluation was performed independently by 2 authors. In other studies, the JBI Critical Appraisal Checklist for Case Series was applied. Because iLGG and sLGG represent different stages of glioma, it is impossible to match 2 groups appropriately. Hence, the assessment of lead-time and length-time bias was additionally performed. The publication bias was evaluated using a funnel plot in the case of more than 5 studies; however, a linear regression analysis was not performed because each evaluation included less than 10 studies.

Ethical Approval and Informed Consent

This review did not involve direct studies on humans, so no informed consent was required.

Results

We retrieved 77 articles out of 298 records identified in databases for full-text assessment (Figure 1). Sixty were excluded; (review articles or commentary on incidental gliomas [n = 25], did not include incidental cases [n = 22], did not include glioma cases [n = 2], lacked necessary data [n = 8], showed overlapping data from institutions of included studies [n = 3]). As a result, we selected 17 articles for data extraction (Supplementary Table 1).^{5–10,15,21–30} In 6 studies from 5 institutes, we compared the OS of iLGG patients to that of sLGG patients (Table 1).^6,7,9–11 Three of the studies were used only to calculate lead-time bias, and another 4 case reports were for the calculation of time before the onset of symptoms. Data on patient characteristics were extracted in 10 articles from 8 institutes. Two studies were from the same institution, and 1 was a multi-institutional study from institutions that had already published articles. Therefore, we selected articles that had the necessary data (patient characteristic data, n = 8; comparison between iLGG and sLGG, n = 5).

Risk of Bias

With the exception of 1 study, all were retrospective case–control studies or case series. All studies had a low evidence level. iLGG and sLGG were different in many points that were evaluated in the next section. The JBI checklist for case–control studies showed at least 2 or 3 deficits (Supplementary File 2). However, we did not exclude the studies because the majority were inevitable due to differences in the size and location of the tumors.

Overview of the Clinical Characteristics of Incidental Gliomas

The main reasons for the initial radiological examination were headaches unrelated to iLGG (23.4%), head injury (19.4%), and screening for other diseases (19.9%) (Supplementary Table 2). The pooled data in iLGG showed that the mean age of the patients with iLGG were 38.6 years (8 studies) with a female preponderance (54.4%) (8 studies) (Table 2) (Forest plots in Supplementary Figure S1). Oligodendrogliomas (8 studies) or gliomas with Codel (7 studies) accounted for more than 50% of cases. iLGGs were associated with a significantly higher frequency of oligodendroglioma histology (pOR 1.59, 95% CI [1.05–2.39], P = .028, 5 studies) and female sex (pOR 1.60, 95% CI [1.26–2.04], P = .0001, 5 studies) than sLGGs. iLGGs were smaller in size (Mean Difference −45.9, 95% CI [−55.5–−36.3], P < .0001, 4 studies) and were less frequently located in eloquent areas (pOR 0.046, 95% CI [0.012–0.174], P < .0001, 3 studies) in comparison to sLGGs. As a result, the formers were more amenable to total resection (OR 2.56, 95% CI [1.30–5.05], I² = 79%, 4 studies) (Figure 2A) The pooled total removal rate was 62.1% (95% CI [49.5–77.8] I² = 95.1%, 8 studies) in iLGG (Figure 2B) The meta-regression analysis revealed that the high heterogeneity was explained by the difference in preoperative volume (Figure 2C). The smaller the preoperative volume, the higher rate of total removal.

The proportion of oligodendroglioma (both iLGGs and sLGGs) was highly heterogeneous among institutes (Table 2). Heterogeneity was also seen in the molecular diagnosis (Table 2) and was correlated with the 10-year OS of sLGG in the meta-regression analysis (Supplementary File 1). We suspected that the heterogeneity was due to regional characteristics but not due to diagnostic misclassification.

The observation time from the diagnosis to surgery in iLGG was variable, the mean time was from 10.4 to 34.4 months (median 3.1–22 months) (Supplementary Table 1). The longest observation period without symptomatic change was 170 months.²²

Estimation and Correction of Lead-Time Bias

We found the LTs value in 4 studies and 5 single case reports.^{7,9,15,25–30} The majority of patients who progressed symptomatic experienced seizures (Supplementary Table 3). The LTs ranged from 1 to 171 months. When 5 case reports were aggregated to one, a leave-one-out test demonstrated that there was no outlier. Hence, the pooled result of the mean LTs was 45.1 months (3.76 years, n = 50) (95% CI [34.8–58.3], I² = 74%) (Figure 3A). The leave-one-out test showed that the pooled mean LTs ranged from 35.4 to 48.0 months. Because of the high heterogeneity, we performed a meta-regression analysis. We found that the year of publication significantly affected heterogeneity (P = .015, estimate −1.538, 95% CI [−2.781 ~ −0.296], I² = 0%). Newer articles tended to have a shorter mean LTs.

When correcting OS by the lead-time bias, LTs should be adjusted by the preoperative observation period, from the diagnosis to surgery, which differed between iLGG and sLGG. However, we did not modify the LTs for the following reasons: In 2 studies, OS was calculated from the time of the radiological diagnosis^7,9; the median observation time from the diagnosis to surgery was 3.1 months in iLGGs but unknown in sLGG in the study by Gogos et al.⁶ but it was 5 months in a study from the same institute that included both iLGG and sLGG³¹; the observation time for either iLGG and sLGG was unknown in the studies by Ius et al.¹⁰ and Wang et al.¹¹

We also calculated the lead time using a tumor growth model (LTg). The pooled mean VDE was 3.14mm/year (95% CI [2.64–3.73] I² = 79.9%) (Figure 3B) from 4 studies in 164 patients that were followed for a mean period of 34.0–63.5 months.^15,22,23,26 LTg was not calculated in one study because of a lack of preoperative volume data.¹¹ In 4 other studies, obtained lead times, which varied from 4.16 years to 6.12 years, were slightly longer than the value of the pooled mean LTs (Table 3).

When we corrected the lead-time bias using values from the mean LTs and LTg separately, three studies lost the statistically significant difference in OS between iLGG and sLGG (Table 3, Figure 3CD). However, the pooled HR (pHR) showed significantly better survival in patients with iLGG than those with sLGG (pHR 0.64 [95%CI {0.51–0.81}]) in LTs, and pHR 0.70 (95% CI [0.56–0.88]) in LTg (Table 3). Because the weight of the study by Ius et al.¹⁰ in the forest plots estimated values by entering “0.01” for assumed death, we reanalyzed the data with the exclusion of their study. The pHR non-adjusted, LTs adjusted, and LTg adjusted were 0.30 (95% CI [0.19–0.48]), 0.53 (95% CI [0.33–0.84]), and 0.58 (95% CI [0.35–0.95]), respectively.

Two studies reported that the OS of the iLGG subgroup with total removal was superior to that of the sLGG subgroup with total removal.^6,10 However, the superiority was lost when lead-time bias was corrected by LTs (pHR 0.66, 95% CI [0.35–1.24], P = .20) (Figure 3E). Correction by LTg further decreased the difference (pHR 0.89, 95% CI [0.69–1.14], P = .35).

Length-Time Bias

As oligodendrogliomas were more often observed in iLGGs than in sLGGs, we attempted to correct the bias using the difference in frequency. The exact methodology is described in Supplementary File 1. Briefly, the true relative risk of death from iLGG versus sLGG, independent of the length bias was calculated (φ). For this purpose, we calculated the frequency of oligodendrogliomas in patients with iLGG and sLGG in 5 studies with adjustment of lead-time by LTs (correction 1) and 4 studies with the adjustment of lead-time by LTg (correction 2).

p₁=The probability of cancer death for sLGG.

p₂ = The probability of cancer death for iLGG

Because the exact death rates in patients with iLGG and sLGG were not known, we performed a sensitivity analysis in which the death rate of patients with sLGG changed from 50% to 90% (Supplementary File 1). It is unlikely that the death rate exceeds 90% because the pooled value of 10-year OS in patients with sLGG was 54% (95% CI [42–66], I² = 96.2%) (Supplementary File 1). The length-time bias adjusted HR in correction 1 was at most 0.655 (95% CI[0.520–0.834]). In correction 2, it was 0.736 (95% CI [0.583–0.941]). Hence, patients with iLGG still showed better OS in comparison to those with sLGG after the correction of the lead time and length-time.

We did not adjust the HR due to the different sex ratios, because φ was relatively small (1.004 (p₂/ p₁)).

Progression-Free Survival

The PFS of patients with iLGG varied in the different reports. The median PFS ranged from 29 to 85 months,^7–9,21 while the 10-year PFS rate was 51.65%–69.9 %.^9,10 Two studies compared PFS between iLGG and sLGG. One showed no significant difference (log-rank test, P = .633)⁶ whereas the other showed a statistically significant difference (log-rank test, P = .017).¹¹ The pooled HR did not indicate a significant difference (HR0.53, 95% CI [0.18–1.61], P = .26), even without correction of the lead-time or length-time bias.

Malignant Transformation

Malignant transformation in iLGG was reported in 9.7%–32.7% of the cases during various follow-up periods.^6,7,21,23 The incidence of a clinically defined diagnosis (defined either histologically or radiologically) was higher than that of the histologically proven diagnosis. The pooled incidence of clinically defined malignant transformation was 4.19/ 100 person-years (95% CI [3.11 – 5.65], I² = 0%)^7,8,15,21 whereas a multi-institutional study reported a histologically proven incidence of 0.68/ 100 person-years.²⁴

Gogos et al.⁶ showed no significant difference in MPFS between iLGG and sLGG (log-rank test, P = .25), even without correction of the lead-time bias. Also, Zeng et al.¹⁵ reported no difference in MPFS between iLGGs with symptomatic progression and iLGGs without symptoms.

Neurological Deficits After Surgery in iLGG

One multi-institutional study²⁴ and another study¹⁵ reported transient or permanent deficits after surgery. The pooled incidence was 12.3% (95% CI [1.81–83.7]) for transient deficits and 2.6% (95% CI [1.49–5.62]) for permanent deficits (Table 4). While two studies compared the rate of deficits between iLGG and sLGG,^8,10 neither of the studies showed statistical significance. The pooled risk ratio of iLGG for sLGG for transient deficits was 0.75 (95% CI [0.40–1.43]) and that for permanent deficit was 0.55 (95% CI [0.07–4.19]).

The pooled incidence of early seizure was 9.43% (95% CI [4.85–17.5]) and that of late seizure was 5.24% (95% CI [1.55–16.2]) (Table 4).

Second Search

We performed a second search on March 1, 2022 (Supplementary Fig. S2) and found one new article that compared OS between iLGG (n = 20) and sLGG.³² This article found no significant difference in OS (P = .074), with a maximum follow-up period of 5 years. Because the follow-up time was too short for lead-time correction, we did not include this article in the analysis. This study showed higher odds ratios for oligodendrogliomas (3.12, 95%CI [0.90–10.80]) and female sex (1.16, 95% CI [0.42–3.14]) in patients with iLGG in comparison to patients with sLGG, as shown in this meta-analysis.

Discussion

We demonstrated that the OS of patients with iLGG was significantly longer than that of patients with sLGG, even after correction of lead-time and length-time bias, although it was not as distinct as previously reported in the literature. On the other hand, OS of totally removed iLGG did not differ from that of totally removed sLGG after correction of the biases. Moreover, it was not clear whether PFS and MPFS in iLGG were longer than those in sLGG.

Lead-Time Bias

Lead-time bias is a well-known problem in cancer screening programs. In cancer screening, investigators have to compare mortality rates from a disease randomized to the screening of the whole control population to avoid the bias. Alternative methods have been developed to compare screening-detected cases with symptomatic cases because randomization is not always possible.^12,17,33

We used 2 methods to estimate the lead time. One was using the LTs for the lead time. Because iLGG causes seizure as the first symptom in most cases, the time from the incidental diagnosis to the first symptom is clear. If “wait and see” is a dominant policy, the LTs would accurately reflect the lead time. However, our study raised a question regarding accuracy. The synthesized mean LTs showed high heterogeneity that was related to the year of publication of the study. We suspected that the recent tendency toward early intervention for iLGG decreased the number of cases with long-term follow-up. That would have resulted in a decrease in the LTs. Therefore, the obtained pooled results for the mean LTs (3.76 years) were likely to have been short. Actually, the LTg was longer than the LTs. LTg has advantages over the LTs for statistical assessment. Firstly, the LTg was determined in each study when the preoperative volume was available for both iLGG and sLGG. This partly diminished the bias caused by the difference in tumor volume between iLGG and sLGG. Secondly, the difference in the preoperative observation time between iLGG and sLGG would affect the outcome data when the LTs is used for lead-time correction.

The problem in using LTg in the correction of the lead time is that there is currently no precise growth model for LGG. Although the majority of studies of growth kinetics in LGG used VDE as an index,^20,26,34 some authors insisted on applying an exponential growth model.^6,22 In the VDE model, tumor growth shows a cubic curve. The mVDE was obtained from the mean observation time of 35–63.5 months. Cubic and exponential curves may show little difference in a relatively short period. However, the calculation of the growth time by the former could overestimate the lead time if a tumor grows exponentially, especially when the estimation is outside of the observation time in which the VDE is calculated. The LTg obtained in this study was not likely to have been overestimated by a great deal in the case of exponential growth, since it ranged from 4.16 years to 6.12 years, and stayed mostly within the range of the observation time.

Length-Time Bias

iLGG has been considered to represent an early phase of sLGG.²⁰ However, iLGGs have different clinical features from sLGGs, other than their size and location. In iLGG there was a higher frequency of oligodendroglioma, and iLGG was more frequently found in females. It is well-known oligodendroglioma is associated with a better prognosis than diffuse astrocytoma. Although still controversial, the majority of studies showed better survival of females with LGG than males.^24,35 Therefore, patients with iLGG have a clinically better prognosis than patients with sLGG. Although no articles showed faster growth of sLGG in comparison to iLGG, oligodendroglioma is reported to grow more slowly than astrocytomas.^34,36,37 Moreover, the malignant transformation rate of astrocytoma was higher than that of oligodendroglioma.³ This may be also true for sex. Nevertheless, the correction of length-time bias in this study had little effect on the HR, because of the relatively small difference in the frequency of oligodendroglioma between the iLGG and sLGG (Supplementary File 1).

Treatment of Incidental Low-Grade Gliomas

The OS of patients with iLGG was longer than that of patients with sLGG, even after correction of the biases. However, the difference was not as distinct as previously reported in the relevant literature. One of the reasons was that the iLGG cases had a median observational period of 3.1–21 months, which led to enlargement of the tumors. As a result, cases that had become symptomatic were included in iLGGs in one study.⁷ While 62% of iLGGs were totally removed at the initial surgery (Table 2), the gains in OS in patients with totally removed iLGG in comparison to totally removed sLGG were lost after correction of the lead-bias, even without the correction of the length-time bias. Therefore, the increased total removal rate of iLGG appeared to contribute to the improvement of OS. Whereas total removal was a more important factor for OS irrespective of the presence of symptoms, the benefit of subtotal removal of iLGG was not clear.

The pooled incidence of clinically defined malignant transformation was 4.19/ 100 person-years in iLGG. The value was not largely different from the results of a recent meta-analysis that showed a 10-year malignant transformation rate of approximately 40% in patients with LGG.³ Although the improvement of PFS and MPFS were expected in iLGG, which had a higher rate of total resection in comparison to sLGG, we did not find a significant difference in the pooled results. One possible reason was that majority of patients with iLGG did not undergo radiotherapy or chemotherapy after initial resection. Further studies are needed to draw a conclusion. Another issue is the adjustment of the biases in PFS and MPFS. Because tumor growth or malignant transformation occurs even before symptomatic progression, simple adjustment using the LTs or LTg may not be appropriate.

The postoperative neurological deficits of patients with iLGG and those with sLGG were comparable. This result was unexpected because iLGGs are more frequently located in non-eloquent areas. Awake surgery under monitoring can facilitate safe resection even in tumors near eloquent areas irrespective of whether they are classified as iLGG or sLGG. In contrast, it should be noticed that even asymptomatic patients may have postoperative deficits (pooled incidence 12.3% in transient deficits, 2.6% in permanent deficits). Permanent deficits in early surgery may result in a longer-lifetime burden than a watch-and-wait policy, while it is possible that surgical and imaging advances will make surgery safer in the future.

Clinical Implications

Although our study showed that the early treatment of iLGGs was beneficial, iLGG may not be easy to diagnose. The HUNT MRI study,³⁸ which systematically assessed the prevalence of incidental intracranial findings, reported that only 1 of 13 cases that were initially suspected as LGG was true LGG after additional MRI examination and MR spectroscopy. Although MRI spectroscopy or positron emission topography may be useful for the differential diagnosis, there are no definite methods or techniques. The majority of studies recommended MRI follow-up every 3–6 months,^14,23 and surgical treatment after the confirmation of growth. Boetto et al.³⁹ reported that approximately 18.8% of incidental findings were stable over time among the cases of suspected iLGG that they managed, whereas the histopathological diagnosis of growing lesion was diffuse glioma in all cases. However, there were cases that showed very slow growth, (VDE < 1 mm/year).²⁰ Because such a small change is within the measurement error, we often have to follow up on patients with suspected iLGG for a long time. Although malignant transformation during the observation period may be a concern, smaller lesions or slow-growing lesions are less likely to undergo transformation.^3,26,40 On the other hand, it is important to remove a lesion before it grows beyond a size at which total removal can be safely performed.

It may not be clear whether all iLGGs become symptomatic because reports have been based on surgically treated cases. In our search, however, no iLGGs remained asymptomatic at 171 months after the radiological diagnosis.

Limitations

The majority of studies analyzed in the present study were retrospective and had a low evidence level. However, the rarity of iLGG and the need for long-term follow-up preclude a prospective randomized study. Thus far, a meta-analysis would be a better method for gaining a convincing result from previously reported articles. Careful interpretation is necessary because we did not exclude all possible biases.

We extracted survival data from Kaplan–Meier curves in the figures of the articles. Although we confirmed that original and extracted curves overlapped, there may have been measurement errors within the thickness of the graphical line, 0.1%–0.15% at each point.

Correction of the length-time bias is generally challenging. We applied the method of Duffy et al.,¹² which was originally created for cancer screening programs with large cohorts. In this study, the rate of pure oligodendroglioma was very heterogenous among institutes, for both iLGG and sLGG. Therefore, the result may include a certain degree of error, although the correction of the length-time bias was within a smaller range in comparison to the correction of the lead-time bias. Another method using propensity score matching may be appropriate if individual data are available.

The WHO classification system has been revised.⁴ One multicenter study did not include molecular data.⁷ However, the pooled rate of oligodendrogliomas (53.4%) and Codel (52.7%) in iLGG was not different (Table 2). Furthermore, the pooled OR of oligodendrogliomas in iLGG to sLGG (1.59) was very similar to that of Codel (1.60). Therefore, the effect of change in tumor classification appeared small in this study. IDHw LGG, which accounted for 14.1% of the cases of iLGG in this study may not be classified as LGG according to the new criteria. Because the frequency of IDHw LGG in iLGG and sLGG did not differ substantially (Table 2), it is unlikely that it had any significant impact on the HR.

Conclusions

iLGG is a group of gliomas that progress to become symptomatic but which tend to have slightly better clinical characteristics in comparison to sLGG. The estimated mean lead time was approximately 4–6 years. Although patients with iLGG had longer OS than those with sLGG after correction of the lead-time and length-time biases, the difference was not as distinct as previously reported. The benefit of total removal did not differ between iLGG and sLGG. The improvement of PFS, MPFS, and postoperative neurological deficit in comparison to sLGG has not been demonstrated yet.

Acknowledgments

We would like to thank Mr. Brian Quinn (Japan Medical Communication) for editing a draft of this manuscript.

Funding

No specific grant was received for this research.

Conflict of interest statement. The authors have no conflicts of interest to declare.

Authorship statement. Conception and design: S.N., Y.N., K.N. Data acquisition: S.N., Y.N., A.T., T.F., N.N., H.K. Data analysis. S.N., Y.N., A.N., Data interpretation: S.N., Y.N., A.N., K.N. Manuscript writing: S.N., Y.N., A.T., K.N.

References