ABSTRACT
Objective
The aim of this study is to determine whether a previously reported three‐tiered classification scheme for T2‐hyperintense brain lesions in children with neurofibromatosis type 1 (NF1) was associated with distinct radiographic or clinical characteristics after 10 years of real‐world follow‐up data were obtained.
Methods
In this single‐center retrospective study, we analyzed T2‐hyperintense lesions identified on brain magnetic resonance imaging (MRI) of 58 children with NF1. Lesions were categorized as “Low Likelihood of Tumor,” “Intermediate Likelihood of Tumor,” or “High Likelihood of Tumor” using predefined imaging criteria adopted by our institution. Clinical and radiologic data were evaluated with generalized estimating equations (GEEs).
Results
No demographic variables were associated with lesion category, apart from small age differences and a lower proportion of “High Likelihood of Tumor” lesions occurring in males compared with “Low Likelihood of Tumor.” Localization patterns differed modestly, but anatomic site was not associated with radiographic evolution or treatment. Most lesions initially classified as “Low Likelihood of Tumor” or “Intermediate Likelihood of Tumor” remained stable or resolved, and none required therapy. Even among lesions initially classified as “High Likelihood of Tumor,” only a small subset required treatment, most often those demonstrating pronounced T1 hypointensity.
Interpretation
NF1‐associated T2‐hyperintense lesions demonstrate substantial overlap in clinical and radiographic characteristics, particularly those classified as “Low Likelihood of Tumor” and “Intermediate Likelihood of Tumor,” while generally following a benign course. Features such as lesion location, enhancement, or mass effect are not consistently associated with treatment. These findings support close monitoring of lesions classified as “High Likelihood of Tumor” with pronounced T1 hypointensity.
Keywords: focal areas of signal intensity, glioma, NF1, optic pathway glioma, T2 hyperintensity
1. Introduction
Individuals with the neurofibromatosis type 1 (NF1) neurogenetic syndrome are at a high risk of developing brain tumors throughout their lives [1, 2, 3, 4]. In childhood, these tumors are typically low‐grade gliomas (World Health Organization grade 1 and 2 astrocytomas) that arise within the optic pathway (optic nerves, chiasm, tracts, and radiations) [5, 6], although other areas of the brain can be involved [7, 8, 9]. Optic pathway gliomas (OPGs) are usually detected in younger children with NF1, with a median age at OPG diagnosis of approximately 4 years [10, 11]. While many of these tumors remain radiographically unchanged, are clinically asymptomatic, or do not require treatment, some can cause progressive vision loss or endocrinological complications, including precocious puberty [12, 13].
In addition, over 90% of children with NF1 harbor “bright spots” on T2‐weighted or FLAIR (fluid‐attenuated inversion recovery) magnetic resonance imaging (MRI), referred to as T2 hyperintensities, focal areas of signal intensity (FASI), or unidentified bright objects (UBOs) [14, 15]. While these abnormalities are not thought to be of clinical significance [16], it is often difficult to distinguish these benign T2 hyperintensities from nonoptic pathway low‐grade gliomas [17] —tumors that are more prevalent in NF1 than commonly recognized—underscoring the clinical importance of accurate lesion classification to guide appropriate surveillance and management.
For these reasons, our team previously developed a classification scheme, based on proposed radiologic criteria, separating T2 hyperintensities into three groups: “Low Likelihood of Tumor,” “Intermediate Likelihood of Tumor,” and “High Likelihood of Tumor” [8, 15]. We identified differences in laterality, border definition, location, and age of onset. In addition, we found that many lesions, including those that evolved radiographically, remained clinically benign, while nearly 1/3 of patients with probable tumors required treatment. Following completion of this study, our pediatric neuroradiology division universally adopted this three‐tiered classification system with some minor institutional refinements in an effort to provide uniform interpretations and help guide medical surveillance and treatment planning.
Given that lesions classified as “Intermediate Likelihood of Tumor” were grouped with those classified as “Low Likelihood of Tumor” in the original study (“UBOs”), it remained unclear to the practicing physicians whether these lesions should be considered a clinically distinct entity and thus warrant closer monitoring. To address this, we extended the analysis of all T2‐hyperintense lesions observed in children with NF1 from our original study to include an additional 10 years of clinical data. This new information not only allowed us to compare baseline characteristics across all three likelihood categories but also provided an opportunity to track their clinical progression and determine whether “Intermediate Likelihood” lesions warrant separate surveillance considerations.
2. Methods
2.1. Study Population
This retrospective cohort study included the same patients identified in our previous study [15], all of whom met updated National Institutes of Health (NIH) diagnostic criteria for NF1 [18], were 21 years of age or younger, and had at least one brain MRI performed at St. Louis Children's Hospital (Washington University School of Medicine [WUSM]) between January 1, 2006 and December 31, 2016 (n = 72 subjects). No sample size calculation was performed, given the exploratory nature of the study. Thirteen patients were excluded from the final analysis due to insufficient imaging or clinical follow‐up information, given the institution's electronic medical records transition in 2018. An additional patient was excluded following neuroradiology review because their lesions could not be distinguished from OPG involvement, leaving a total of 58 participants. With the approval of the WUSM institutional review board (IRB), data from their electronic medical records were extracted through May 7, 2026, comprising age, biological sex, race/ethnicity, family history of NF1, tumor diagnoses, pathology results, treatment history, brain MRI scans, their corresponding radiology reports, and relevant clinical documentation to assess lesion‐directed management. All information was coded and stored in a HIPAA‐compliant REDCap database.
2.2. Assessment of T2 Hyperintensities
All patients with ≥ 1 MRI were included in baseline lesion analysis to compare demographic and localization patterns among the different lesion categories. Available MRI scans performed at St. Louis Children's Hospital between January 1, 2006, and December 31, 2016, were analyzed as part of the previous study by J. L. G., with training and arbitration from an experienced neuroradiologist involved in the original study (M. S. G.) [15]. Using standardized protocols, T2‐hyperintense lesions detected on FLAIR and T2‐weighted sequences were characterized by location, border (discrete or nondiscrete), shape (round/ovoid, complex, or multilobular), T1 hypointensity (mild = hypointensity on MPRAGE (magnetization‐prepared rapid gradient echo) sequences; moderate = hypointensity on T1 spin‐echo sequences; and severe = more hypointensity than gray matter on any T1‐weighted sequence), mass effect (yes or no), contrast enhancement (faint, heterogeneous, or homogeneous), and interval change in size (if > 1 MRI was available for analysis).
Lesions identified in the previous study were then classified using the previously established three‐tiered framework [15]. “High Likelihood of Tumor” was defined as a lesion showing either (1) mass effect/architectural distortion or (2) enhancement. A lesion was considered “Intermediate Likelihood of Tumor” if it showed (1) discrete borders of T2 hyperintensity, (2) T1 hypointensity relative to the nearby gray matter, or (3) interval growth in the teenage years. “Low Likelihood of Tumor” was identified as being ill‐defined T2 hyperintensities without associated T1 hypointensity, mass effect, or interval growth in the teenage years. These correspond to the lesion categories used in the original study [15], with two minor institutional refinements. First, our institution now considers severe T1 hypointensity as a criterion for “Intermediate Likelihood of Tumor” rather than “High Likelihood of Tumor,” given that 92% of “High Likelihood” lesions in our previous study were classified based on enhancement or mass effect rather than T1 hypointensity alone [15]. Second, interval growth in the teenage years at our institution is now considered an additional criterion for “Intermediate Likelihood,” given that T2 hyperintensities typically regress during the second decade of life [14]. Representative radiographic images of lesions using this classification scheme are depicted in Figure 1.
Figure 1.
Classification scheme used to define T2‐hyperintense lesions. Schematic outlining the imaging characteristics used to classify T2‐hyperintense lesions as “Low Likelihood of Tumor,” “Intermediate Likelihood of Tumor” or “High Likelihood of Tumor” in this study. Representative magnetic resonance images are shown to illustrate key features associated with each classification. Arrows denote the lesions.
2.3. Longitudinal Evaluation of Lesions
Longitudinal analyses were restricted to patients with ≥ 2 MRI scans in the initial study. Furthermore, an individual lesion had to be present on more than one MRI from the previous study to evaluate interval changes in radiographic features and classification. Specific attention was paid to lesions initially classified as “Low Likelihood of Tumor” or “Intermediate Likelihood of Tumor” to identify potential factors associated with evolution versus stability or resolution. Lesions were matched across time points according to anatomical location. These lesions were then grouped into the following categories: “Stable” if the lesion remained unchanged in classification or was downgraded to a less concerning category (e.g., a lesion initially classified as “Intermediate Likelihood of Tumor” that remained “Intermediate Likelihood of Tumor” or became “Low Likelihood of Tumor”), “Evolving” if the lesion was reassigned to a category that was more concerning (e.g., a lesion initially classified as “Low Likelihood of Tumor” that was later reclassified as “Intermediate Likelihood of Tumor” or “High Likelihood of Tumor”), or “Resolved” if the lesion was no longer visible on the final MRI. Once these lesions were grouped based on their radiographic appearance, a manual review of the electronic medical records was performed to provide clinical correlations, specifically neurological signs, symptoms, or interventions resulting from a particular lesion. Radiology reports in the electronic medical records were used to align clinical documentation with previously identified lesions.
2.4. Treatment Analysis
For lesions initially classified as “High Likelihood of Tumor,” treatment status was determined by manual review of the electronic medical records and radiology reports. A lesion was categorized as “Treated” if the patient received surgery, chemotherapy, or radiotherapy specifically directed at that lesion. Otherwise, a lesion was considered “Untreated” if no lesion‐specific therapy was administered. Imaging characteristics at the time of initial detection of these “Treated” and “Untreated” lesions were then compared. When there was uncertainty about whether a lesion was contiguous with an OPG, rather than an independent lesion, the MRI images containing the relevant lesions were independently reviewed by an expert neuroradiologist involved in the original study (M. S. G.) to confirm spatial independence from the optic pathway.
2.5. Bias
This study used MRI data previously collected using a standardized clinical protocol outlined in our previous study [15]. Lesions were classified using predefined radiographic criteria applied uniformly across all cases to minimize classification bias. Other potential sources of bias were limited by using a single‐institution cohort, uniform imaging acquisition parameters, and consistent reviewer training.
2.6. Statistical Analysis
All analyses were conducted in RStudio (version 4.5.1). There were no missing data for the variables included in the analyses. Because some patients contributed multiple lesions, each lesion was treated as an individual observation, and generalized estimating equations (GEEs) with an exchangeable correlation structure were used to account for within‐patient clustering. Quantitative variables were analyzed as continuous variables without additional categorization. Population‐averaged odds ratios (ORs) with 95% confidence intervals (CIs) were reported. If a GEE model failed to converge due to sparse data or minimal outcome variability, we refit the model using a simpler independence correlation structure or a logistic regression model with cluster‐robust standard errors to maintain valid inference when necessary. Statistical significance was defined as two‐tailed p < .05.
2.7. Standard Protocol Approvals, Registrations, and Patient Consents
This retrospective study was approved by the Washington University School of Medicine Institutional Review Board (IRB). The requirement for informed consent was waived due to the minimal‐risk nature of the study and use of de‐identified clinical data.
3. Results
3.1. Study Population and Lesion Characteristics
A total of 72 children with NF1 were identified in the original study based on updated NIH diagnostic criteria [18]. After further review, 13 children were excluded due to incomplete imaging or insufficient clinical information, while 1 child was excluded because their lesions could not be confirmed as spatially independent from the optic pathway tumor, leaving 58 patients for subsequent analysis. Of these subjects, 28 (48%) were female, and the majority were White (86%). The mean age at initial MRI was approximately 6 years. A concurrent OPG was present in 18 patients (31%). Thirty‐nine patients (67%) had more than one MRI available for review, with a median of two MRI scans available per patient. These data are included in Supporting Information S1: Table 1.
Fifty‐four patients contributed 403 unique T2‐hyperintense lesions for analysis: 228 lesions (57%) were classified as “Low Likelihood of Tumor,” 145 (36%) as “Intermediate Likelihood of Tumor,” and 30 (7.4%) as “High Likelihood of Tumor.” While four patients did not contribute any lesions to the overall analysis, it should be appreciated that most patients harbored more than one lesion type. In this regard, 84% had at least one lesion classified as “Low Likelihood of Tumor,” 78% had at least one lesion classified as “Intermediate Likelihood of Tumor,” and 40% had at least one lesion classified as “High Likelihood of Tumor” (Table 1). There was no significant difference in race/ethnicity across lesion classification categories, and the presence of an OPG was not associated with lesion classification (Supporting Information S2: Table 2). However, lesions classified as “High Likelihood of Tumor” were less likely to occur in male patients than those classified as “Low Likelihood of Tumor” (“High” vs. “Low”: OR 0.36, 95% CI: 0.15‐0.82, p = .015), although there were no significant differences in sex between “Intermediate Likelihood of Tumor” and “Low Likelihood of Tumor” or “High Likelihood of Tumor.”
Table 1.
Demographics by lesion type.
| Characteristics | “Low” | “Intermediate” | “High” |
|---|---|---|---|
| Number of patients with specified MRI abnormalities (n = 58; %) | 49 (84) | 45 (78) | 23 (40) |
| Sex (% females with specified MRI abnormalities) | 26 (53) | 23 (51) | 14 (61)a |
| Race/ethnicity (% White with specified MRI abnormalities | 45 (92) | 40 (89) | 21 (91) |
| Mean age lesion first appeared on MRI, y (range) | 6.1 (1–14) | 7 (2–16)a | 7.4 (3–14)a |
| Concurrent OPG (% with specified MRI abnormalities) | 18 (37) | 17 (38) | 8 (35) |
Note: Refer to Supporting Information S2: Table 2 for full statistical results.
Statistically significant difference compared with “Low Likelihood of Tumor” (p < .05).
Lesions classified as “Intermediate Likelihood of Tumor” and “High Likelihood of Tumor” were detected at slightly older ages relative to “Low Likelihood of Tumor” (“Intermediate” vs. “Low”: OR 1.11, 95% CI: 1.02‐1.21, p = .016; “High” vs “Low”: OR 1.2, 95% CI: 1.06‐1.35, p = .004), suggesting that lesions with more concerning features tend to emerge later in childhood. However, these differences were small in magnitude, and age did not differ significantly between “High Likelihood of Tumor” and “Intermediate Likelihood of Tumor” lesions.
With respect to brain localization, lesions classified as “Low Likelihood of Tumor” most commonly involved the medial temporal lobe (27%), the thalamus (21%), and the brainstem (19%), with additional involvement of the cerebellum (16%) and less frequent localization to the basal ganglia (10%) and other regions (7%) (Table 2). Those lesions classified as “Intermediate Likelihood of Tumor” were most frequently located in the cerebellum (33%) but also had a relatively high prevalence in the brainstem (28%) and the basal ganglia (26%), while thalamic (5%) and medial temporal lobe (1%) involvement was uncommon. Lesions classified as “High Likelihood of Tumor” had the highest frequency in the brainstem (47%), followed by the basal ganglia (30%), with less frequent involvement of the cerebellum (10%), the medial temporal lobe (7%), and the thalamus (3%).
Table 2.
Prevalence of lesions based on brain location.
| Location (n = number of lesions) | “Low” (n = 228) | “Intermediate” (n = 145) | “High” (n = 30) |
|---|---|---|---|
| Medial temporal lobe (%) | 61 (27) | 2 (1)a | 2 (7)a |
| Thalamus (%) | 49 (21) | 7 (5)a | 1 (3)a |
| Brainstem (%) | 43 (19) | 41 (28) | 14 (47) |
| Cerebellum (%) | 36 (16) | 48 (33) | 3 (10)b |
| Basal ganglia (%) | 23 (10) | 38 (26) | 9 (30) |
| Other (%)c | 16 (7) | 9 (6) | 1 (3) |
Note: Refer to Supporting Information S3: Table 3 for full statistical results.
Statistically significant difference compared with “Low Likelihood of Tumor” (p < .05).
Statistically significant difference compared with “Intermediate Likelihood of Tumor” (p < .05).
Other includes subcortical white matter, corpus callosum, and internal capsule.
Statistical comparison of localization patterns revealed differences between “Low Likelihood of Tumor” lesions and those classified as “Intermediate Likelihood of Tumor” or “High Likelihood of Tumor” (Supporting Information S3: Table 3). Lesions classified as “Intermediate Likelihood of Tumor” or “High Likelihood of Tumor” were less likely to involve the medial temporal lobe (“Intermediate” vs. “Low”: OR 0.03, 95% CI: 0–0.26, p = .001; “High” vs. “Low”: OR 0.10, 95% CI: 0.02‐0.57, p = .010) or the thalamus (“Intermediate” vs. “Low”: OR 0.14, 95% CI: 0.05–0.43, p < .001; “High” vs. “Low”: OR 0.06, 95% CI: 0.01‐0.51, p = .010). Between lesions classified as “High Likelihood of Tumor” and those classified as “Intermediate Likelihood of Tumor,” “High Likelihood of Tumor” lesions were less likely to be found in the cerebellum (OR 0.17, 95% CI: 0.03–0.83, p = .029).
3.2. Clinical Outcomes
Of the 403 lesions identified, 252 (61%) were present on more than one MRI scan (Figure 2), allowing for assessment of lesion evolution. Of these 252 lesions, 155 were initially classified as “Low Likelihood of Tumor” (61%), 81 were initially classified as “Intermediate Likelihood of Tumor” (32%), and 16 were initially classified as “High Likelihood of Tumor” (6%). Among the lesions initially classified as “Low Likelihood of Tumor,” the majority either remained classified as “Low Likelihood of Tumor” (59%) or resolved spontaneously over time (28%), while only a small proportion radiographically evolved to “Intermediate Likelihood of Tumor” (11%) or “High Likelihood of Tumor” (2%). Similarly, most lesions initially classified as “Intermediate Likelihood of Tumor” remained classified as “Intermediate Likelihood of Tumor” (61%), were downgraded to “Low Likelihood of Tumor” (16%), or resolved (17%), with only 6% evolving radiographically to “High Likelihood of Tumor.” Lesions classified as “High Likelihood of Tumor” were more likely than other lesion types to persist without reclassification (69%), although a minority showed downgrading to “Intermediate Likelihood of Tumor” (12%) or “Low Likelihood of Tumor” (19%) over time. Notably, no lesion initially classified as “High Likelihood of Tumor” resolved spontaneously.
Figure 2.
Evolution of lesions from the initial to the final MRI scan. Pie charts illustrating the final outcomes of lesions initially classified as “Low Likelihood of Tumor” (n = 155), “Intermediate Likelihood of Tumor” (n = 81), and “High Likelihood of Tumor” (n = 16). Percentages indicate the proportion of lesions within each initial classification according to their final MRI outcome. Gray indicates resolved lesions, blue denotes “Low Likelihood of Tumor,” yellow denotes “Intermediate Likelihood of Tumor,” and orange denotes “High Likelihood of Tumor.”
Lesions initially classified as “Low Likelihood of Tumor” or “Intermediate Likelihood of Tumor” were further stratified based on their subsequent clinical behavior (Table 3). As expected, based on our predefined classification criteria, lesions identified as “Evolving” were more likely to develop T1 hypointensity, become expansile, or show contrast enhancement over time relative to “Stable” or “Resolved” lesions (Supporting Information S4: Table 4), consistent with the imaging features used to define radiographic evolution. “Evolving” lesions were more likely than “Resolved” lesions to be located in the basal ganglia (OR: 30.78, 95% CI: 1.72–552.33, p = .020), the cerebellum (OR: 13.29, 95% CI: 1.23–143.76, p = .033), the thalamus (OR: 67.02, 95% CI: 3.87‐1161.29, p = .004), or other locations (OR: 24.15, 95% CI: 3.19–182.79, p = .002) compared with brainstem lesions, though confidence intervals were wide, given the small numbers in each location. Otherwise, “Evolving” lesions did not differ from “Resolved” or “Stable” lesions with respect to shape change or age at first appearance. “Resolved” lesions were less likely to change shape relative to “Stable” lesions (OR: 0.23, 95% CI: 0.06–0.85, p = .028) or be located in the thalamus (OR: 0.18, 95% CI: 0.05‐0.57, p = .004), but no additional differences were observed between “Resolved” and “Stable” lesions. Importantly, no lesion initially classified as “Low Likelihood of Tumor” or “Intermediate Likelihood of Tumor,” even those that evolved radiographically, was treated with lesion‐targeted therapy.
Table 3.
Characteristics of low and intermediate likelihood lesions according to outcome.
| Characteristics (n = number of lesions) | “Evolving” (n = 26) | “Stable” (n = 153) | “Resolved” (n = 57) |
|---|---|---|---|
| Mean age at which lesion was first seen on MRI, y (range) | 4.5 (1–9) | 6.5 (1–14) | 5.0 (1–10) |
| Mean age at reclassification or resolution, y (range) | 7.4 (4–12) | N/A | 8.6 (2–16) |
| T1 hypointensity (%)a | |||
| Decreased | 1 (4) | 21 (14) | 2 (4) |
| Increased | 19 (73)b , c | 6 (4) | 5 (9) |
| Size (%) | |||
| Smaller | 1 (4) | 16 (10) | 4 (7) |
| Bigger | 19 (73) | 17 (11) | 1 (2) |
| Became expansile (%) | 7 (27)b , c | N/A | 0 |
| Enhancement change (%) | |||
| Increased (none→ faint) | 2 (8)b , c | N/A | 0 |
| Border change (%) | |||
| Discrete→ nondiscrete | 1 (4) | 4 (3) | 0 |
| Nondiscrete→ discrete | 8 (31) | 3 (2) | 2 (4) |
| Shape change (%) | 7 (27) | 33 (22) | 4 (7)b |
| Location (%) | |||
| Basal ganglia | 10 (38)c | 22 (14) | 5 (9) |
| Cerebellum | 6 (23)c | 36 (24) | 13 (23) |
| Brainstem | 5 (19) | 27 (18) | 20 (35) |
| Thalamus | 3 (12)c | 30 (20) | 5 (9)b |
| Medial temporal lobe | 0 | 29 (19) | 10 (18) |
| Otherd | 2 (8)c | 9 (6) | 4 (7) |
| Treated (%) | 0 | 0 | 0 |
Note: Refer to Supporting Information S4: Table 4 for full statistical results.
T1 hypointensity grades are defined relative to nearby gray matter.
Statistically significant difference compared with “Stable” lesions (p < .05).
Statistically significant difference compared with “Resolved” lesions (p < .05).
Other includes subcortical white matter, corpus callosum, and internal capsule.
To determine whether treated lesions had ever demonstrated characteristics of “Low Likelihood of Tumor” or “Intermediate Likelihood of Tumor” at any previous imaging time point, we performed a separate analysis adding two more NF1 patients who harbored lesions requiring intervention and were not included in the original study. These MR images were reviewed by an experienced neuroradiologist involved in the original report (M. S. G.). Of the three lesions analyzed, two demonstrated features consistent with “High Likelihood of Tumor”—specifically contrast enhancement and/or mass effect—from the earliest available imaging time point. The remaining treated lesion was initially classified as “Intermediate Likelihood of Tumor,” based on mild T1 hypointensity and subsequent interval growth in the teenage years, before ultimately developing mass effect and meeting criteria for “High Likelihood of Tumor” at the time of biopsy, when it was confirmed to be a low‐grade glioma (Figure 3). Notably, interval growth in the teenage years was the only criterion for “Intermediate Likelihood of Tumor” that was not observed among lesions that evolved radiographically in the primary analysis.
Figure 3.
Longitudinal imaging of a biopsy‐confirmed low‐grade glioma initially classified as “Intermediate Likelihood of Tumor.” Magnetic resonance images from 2017 and 2023, with the corresponding T2 (left) and T1 (right) images. The arrows denote the lesion of interest, which demonstrated mild T1 hypointensity consistent with “Intermediate Likelihood of Tumor” in 2017, but developed mass effect by 2023, meeting criteria for “High Likelihood of Tumor” at the time of biopsy.
Lastly, lesions initially classified as “High Likelihood of Tumor” were further evaluated based on treatment status, regardless of the number of MRIs available (Table 4). Of the 30 lesions classified as “High Likelihood of Tumor,” only a small subset of lesions (13%) were treated, while the remainder did not receive lesion‐specific therapy. Given the small number of treated lesions, location categories were collapsed into brainstem, cerebellum, and supratentorial location for statistical modeling; the full expanded location breakdown is available in Supporting Information S5: Table 5. Moderate or severe T1 hypointensity was associated with treatment (OR: 25.49, 95% CI: 1.98–327.67, p = 0.013), although the confidence interval was wide. No other imaging features, including age at MRI detection, border, enhancement, expansile morphology, shape, or lesion location, differed between “Treated” and “Untreated” high likelihood lesions (Supporting Information S6: Table 6).
Table 4.
Characteristics of treated vs untreated high likelihood lesions.
| Characteristics (n = number of lesions) | “Treated” (n = 4) | “Untreated” (n = 26) |
|---|---|---|
| Mean age at which lesion was first detected on MRI, y (range) | 8.5 (6–14) | 7.2 (3–14) |
| Mean age at treatment, y (range) | 9.8 (6–14) | N/A |
| Discrete border (%) | 2 (50) | 8 (31) |
| Round shape (%) | 2 (50) | 23 (88) |
| T1 hypointensitya (%) | ||
| None/mild | 2 (50) | 20 (77) |
| Moderate/severe | 2 (50)b | 6 (23) |
| Enhancement (%) | ||
| None | 2 (50) | 15 (58) |
| Faint | 0 | 8 (31) |
| Homogeneous | 0 | 2 (8) |
| Heterogeneous | 2 (50) | 1 (4) |
| Expansile (%) | 3 (75) | 18 (69) |
| Location (%) | ||
| Brainstem | 2 (50) | 12 (46) |
| Supratentorialc | 1 (25) | 12 (46) |
| Cerebellum | 1 (25) | 2 (8) |
Note: Refer to Supporting Information S5: Table 5 for fully expanded characteristic categories. Refer to Supporting Information S6: Table 6 for full statistical results.
T1 hypointensity grades are defined relative to nearby gray matter.
Statistically significant difference compared with “Untreated” high likelihood lesions (p < 0.05).
Supratentorial includes the basal ganglia, the medial temporal lobe, the thalamus, subcortical white matter, the corpus callosum, and the internal capsule.
4. Discussion
In this retrospective cohort study, extending our original analysis with 10 years of additional real‐world follow‐up, we compared the radiographic and clinical characteristics of T2‐hyperintense lesions classified as “Low Likelihood of Tumor,” “Intermediate Likelihood of Tumor,” or “High Likelihood of Tumor.” After accounting for intrapatient clustering, we found minimal demographic differences among lesion categories. The only minor exceptions were a greater proportion of “High Likelihood of Tumor” lesions occurring in female patients compared to “Low Likelihood of Tumor” and small age differences, which were significant but likely reflected the timing of MRI acquisition in this retrospective design rather than biological variation. Although localization patterns differed—particularly between lesions classified as “Low Likelihood of Tumor” and those classified as “Intermediate Likelihood of Tumor” or “High Likelihood of Tumor”—these distinctions did not appear to be associated with differences in clinical course. Most lesions not initially classified as “High Likelihood of Tumor” remained stable or resolved spontaneously, and none required lesion‐directed therapy. Even among lesions initially classified as “High Likelihood of Tumor,” only a small subset received treatment. After 10 years of additional follow‐up, this study highlights three key points: (1) “Low and “Intermediate Likelihood” lesions remain clinically benign, regardless of demographics, brain location, or radiographic progression; (2) lesions demonstrating interval growth in the teenage years are an important subset of intermediate likelihood lesions that may warrant closer clinical monitoring; and (3) treatment is rare even among high likelihood lesions, more often in those with more pronounced T1 hypointensity.
Previous studies have compared the age at appearance of benign‐appearing T2‐hyperintensities and more concerning lesions in children with NF1, showing that lesions with features suggestive of malignancy tend to emerge later in childhood [15, 17]. Our findings align with these reports, although the magnitude of age difference was small and might be influenced more by the timing of initial MRI, approximately 6 years of age in our cohort, than by true biological variation. While previous work has described general NF1 cohort demographics, such as sex distribution, no published series has directly compared demographic characteristics among patients who develop predominantly one lesion category versus another. This is likely because of the high frequency of mixed lesion subtypes within individual patients, which complicates stratification by lesion type. We used GEE to account for intrapatient clustering and observed that “High Likelihood of Tumor” lesions were more likely to occur in females compared with “Low Likelihood of Tumor” lesions, though the biological basis for this observation is unclear and this finding should be interpreted with caution, given the small number of “High Likelihood of Tumor” lesions in our cohort. Whether sex truly represents a meaningful demographic risk factor for developing “High Likelihood” lesions warrants investigation in larger cohorts.
Spatial distribution patterns were also similar among the different lesion categories, with some modest differences between groups, providing additional nuance to previous NF1 imaging reports. Our previous work in the same institutional NF1 cohort essentially used the same three‐tiered classification system although ultimately grouped “Low Likelihood” and “Intermediate Likelihood” lesions together as “UBOs,” and found that UBOs were most common in the cerebellum, the medial temporal lobe, and the thalamus, whereas “probable tumors” were more common in the brainstem and the basal ganglia [15]. For the current study, we used the original three‐tiered classification system without condensing categories to better characterize “Intermediate Likelihood” lesions. “Low Likelihood of Tumor” and “Intermediate Likelihood of Tumor” lesions showed modest, but statistically significant, differences in distribution, with “Intermediate Likelihood of Tumor” lesions being less likely to involve the medial temporal lobe or the thalamus. “High Likelihood of Tumor” lesions were also less likely to localize to these regions relative to “Low Likelihood of Tumor” lesions, but they were also less likely to involve the cerebellum than lesions classified as “Intermediate Likelihood of Tumor.” While our results partially align with patterns identified in our previous work [15], they suggest that when UBOs are considered separately as “Low Likelihood of Tumor” or “Intermediate Likelihood of Tumor,” there is substantial overlap in anatomic distribution between less concerning lesions and more suspicious lesions, rather than distinct localization patterns. Collectively, these findings confirm that location is not a distinguishing feature among lesion categories.
Consistent with previous observations that age and lesion location do not predict tumor progression in NF1 [19], neither demographics nor anatomic site in our cohort provided meaningful insight into lesion behavior. We therefore examined longitudinal outcomes across lesion categories to identify imaging features that might correlate with lesion evolution and treatment decisions. Most lesions categorized as “Low Likelihood of Tumor” or “Intermediate Likelihood of Tumor” remained stable or resolved over time, and radiographic evolution to “High Likelihood of Tumor” was rare. These results confirm the benign natural history of most T2‐hyperintense lesions observed in patients with NF1, as demonstrated in previous longitudinal studies [15, 17, 20, 21, 22]. Similar to our approach, previous studies using three risk‐based lesion categories (“low‐risk,” “medium‐risk,” and “high‐risk” tumor lesions) observed no evolution among the lower‐risk groups [14]. While we did observe rare cases of radiographic evolution, none of these lesions required treatment.
Given these findings, we performed a separate analysis to determine whether treated lesions had ever demonstrated features consistent with “Low Likelihood of Tumor” or “Intermediate Likelihood of Tumor.” Two of the three treated lesions demonstrated features consistent with “High Likelihood of Tumor” from the earliest available imaging time point. Interestingly, one biopsy‐confirmed low‐grade glioma was initially classified as “Intermediate Likelihood of Tumor” based on mild T1 hypointensity and interval growth in the teenage years before ultimately developing mass effect, thereby meeting criteria for “High Likelihood of Tumor” by the year the lesion was biopsied. Interval growth in the teenage years, when T2 hyperintensities typically decrease in number and size, was not observed in any “Intermediate Likelihood of Tumor” lesion that evolved radiographically in the primary analysis, suggesting that it may carry particular clinical significance and warrant close surveillance in this age group.
We also examined lesions initially classified as “High Likelihood of Tumor” to identify characteristics that reliably distinguish lesions that will ultimately require treatment. The only significant difference observed between treated and untreated lesions in our study was more pronounced T1 hypointensity at presentation. Severe T1 hypointensity was previously included as a standalone criterion for “High Likelihood of Tumor,” though the majority of “High Likelihood” lesions in the original study were classified based on mass effect or contrast enhancement [15], so this criterion was subsequently incorporated into “Intermediate Likelihood of Tumor.” The present study suggests that while severe T1 hypointensity may lack specificity as a standalone feature, it retains prognostic value in the presence of mass effect or enhancement. However, most “High Likelihood” lesions in our cohort were managed conservatively, which is consistent with findings from others, who similarly observed that the majority of lesions with enhancement and/or mass effect in adults with NF1 followed an indolent course without intervention, suggesting that the largely conservative history of “High Likelihood” lesions may persist across the lifespan [23].
This retrospective, single‐center design limits generalizability, and limited demographic diversity constrained subgroup analysis. The classification framework, while institutionally adopted, has not been validated in independent cohorts. This is challenging to accomplish because histopathologic confirmation is unavailable for most lesions, reflecting standard NF1 practice in which biopsy is rarely indicated. For this reason, we rely on features such as contrast enhancement and mass effect to define lesions as “High Likelihood of Tumor”; however, these have limited specificity in NF1 and may be observed in non‐neoplastic lesions [15, 23], which may result in overestimation of tumor burden. Furthermore, MRI surveillance intervals varied, and the age at initial imaging may have influenced apparent lesion onset. Although GEE was used to mitigate clustering bias, residual intrapatient correlation cannot be excluded. It is also important to note that treatment analyses were exploratory and limited by small numbers. Strengths include longitudinal follow‐up at a single institution, use of a standardized imaging protocol and classification method, and neuroradiologist oversight. Broader, collaborative studies will be essential to refine classification frameworks and identify imaging or molecular markers that distinguish indolent from clinically significant lesions.
Collectively, these findings suggest that children with NF1 who harbor T2‐hyperintense brain lesions classified as “Low Likelihood of Tumor” or “Intermediate Likelihood of Tumor” can be managed conservatively without immediate intervention, even with radiographic evolution. However, close attention should be paid to lesions that demonstrate growth in the teenage years. In lesions classified as “High Likelihood of Tumor,” the presence of pronounced T1 hypointensity may help identify patients most likely requiring treatment, informing more targeted surveillance strategies for this population. Future multi‐institutional studies using these criteria will be helpful to establish uniform guidelines for the management of children with NF1.
Author Contributions
Ariel B. Brickler: writing – original draft, investigation, data curation, formal analysis, visualization, writing – review and editing. Manu S. Goyal: validation, writing – review and editing, supervision, formal analysis, methodology. Robert C. McKinstry: conceptualization, writing – review and editing, supervision. David H. Gutmann: conceptualization, writing – review and editing, formal analysis, visualization, methodology, supervision. Jennifer L. Griffith: conceptualization, writing – review and editing, supervision.
Ethics Statement
This study was approved by the Institutional Review Board at the Washington University School of Medicine.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Supporting File 1
Supporting File 2
Supporting File 3
Supporting File 4
Supporting File 5
Supporting File 6
Acknowledgments
We thank Dr. Lei Liu and Ruiwen Zhou from the Department of Biostatistics for reviewing the statistical analyses. This work was supported in part by an unrestricted gift from Schnuck Markets Inc.
Data Availability Statement
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions. Additional supporting information can be found online in the Supporting Information section at the end of this article.
References
- 1. Packer R. J., Iavarone A., Jones D. T. W., et al., “Implications of New Understandings of Gliomas in Children and Adults With NF1: Report of a Consensus Conference,” Neuro‐Oncology 22, no. 6 (2020): 773–784, 10.1093/neuonc/noaa036. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Nix J. S., Blakeley J., and Rodriguez F. J., “An Update on the Central Nervous System Manifestations of Neurofibromatosis Type 1,” Acta Neuropathologica 139, no. 4 (2020): 625–641, 10.1007/s00401-019-02002-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Park E. R., Smith K. B., Merker V. L., et al., “Examining Perceived Cancer Risk Among Patients With Neurofibromatosis Type 1,” Journal of Neuro‐Oncology 122, no. 1 (2015): 127–133, 10.1007/s11060-014-1689-6. [DOI] [PubMed] [Google Scholar]
- 4. Gutmann D. H., Rasmussen S. A., Wolkenstein P., et al., “Gliomas Presenting After Age 10 in Individuals With Neurofibromatosis Type 1 (NF1),” Neurology 59, no. 5 (2002): 759–761, 10.1212/wnl.59.5.759. [DOI] [PubMed] [Google Scholar]
- 5. Kotch C., de Blank P., Gutmann D. H., and Fisher M. J., “Low‐Grade Glioma in Children With Neurofibromatosis Type 1: Surveillance, Treatment Indications, Management, and Future Directions,” Child's Nervous System 40, no. 10 (2024): 3241–3250, 10.1007/s00381-024-06430-8. [DOI] [PubMed] [Google Scholar]
- 6. Peltonen S., Kallionpää R. A., Rantanen M., et al., “Pediatric Malignancies in Neurofibromatosis Type 1: A Population‐Based Cohort Study,” International Journal of Cancer 145, no. 11 (2019): 2926–2932, 10.1002/ijc.32187. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Sellmer L., Farschtschi S., Marangoni M., et al., “Non‐Optic Glioma in Adults and Children With Neurofibromatosis 1,” Orphanet Journal of Rare Diseases 12, no. 1 (February 2017): 34, 10.1186/s13023-017-0588-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Mahdi J., Shah A. C., Sato A., et al., “A Multi‐Institutional Study of Brainstem Gliomas in Children With Neurofibromatosis Type 1,” Neurology 88, no. 16 (2017): 1584–1589, 10.1212/WNL.0000000000003881. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Guillamo J. S., Creange A., Kalifa C., et al., “Prognostic Factors of CNS Tumours in Neurofibromatosis 1 (NF1): A Retrospective Study of 104 Patients,” Brain 126, no. 1 (2003): 152–160, 10.1093/brain/awg016. [DOI] [PubMed] [Google Scholar]
- 10. Listernick R., Charrow J., Greenwald M. J., and Esterly N. B., “Optic Gliomas in Children With Neurofibromatosis Type 1,” Journal of Pediatrics 114, no. 5 (1989): 788–792, 10.1016/s0022-3476(89)80137-4. [DOI] [PubMed] [Google Scholar]
- 11. Listernick R., Charrow J., Greenwald M., and Mets M., “Natural History of Optic Pathway Tumors in Children With Neurofibromatosis Type 1: A Longitudinal Study,” Journal of Pediatrics 125, no. 1 (1994): 63–66, 10.1016/s0022-3476(94)70122-9. [DOI] [PubMed] [Google Scholar]
- 12. Katz J., Ratnam S., Listernick R. H., Habiby R. L., and Gutmann D. H., “Precocious Puberty in Children With Neurofibromatosis Type 1,” Journal of Pediatrics 278 (2025): 114440, 10.1016/j.jpeds.2024.114440. [DOI] [PubMed] [Google Scholar]
- 13. Fisher M. J., Avery R. A., Allen J. C., et al., “Functional Outcome Measures for NF1‐Associated Optic Pathway Glioma Clinical Trials,” supplement, Neurology 81, no. S21 (2013): 15–24, 10.1212/01.wnl.0000435745.95155.b8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Han Y., Wang H., and Huang Y., “T2 Hyperintensities in Children With Neurofibromatosis Type 1,” World Neurosurgery 192 (2024): e480–e485, 10.1016/j.wneu.2024.10.001. [DOI] [PubMed] [Google Scholar]
- 15. Griffith J. L., Morris S. M., Mahdi J., Goyal M. S., Hershey T., and Gutmann D. H., “Increased Prevalence of Brain Tumors Classified as T2 Hyperintensities in Neurofibromatosis 1,” Neurology: Clinical Practice 8, no. 4 (2018): 283–291, 10.1212/CPJ.0000000000000494. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Kraut M. A., Gerring J. P., Cooper K. L., Thompson R. E., Denckla M. B., and Kaufmann W. E., “Longitudinal Evolution of Unidentified Bright Objects in Children With Neurofibromatosis‐1,” American Journal of Medical Genetics, Part A 129A, no. 2 (2004): 113–119, 10.1002/ajmg.a.20656. [DOI] [PubMed] [Google Scholar]
- 17. Viguier A., Estublier B., Bertozzi A. I., et al., “Distinguishing Atypical Focal Areas of Signal Intensity From Probable Low‐Grade Gliomas in the Posterior Fossa of Children With Neurofibromatosis Type 1,” Archives de pédiatrie: organe officiel de la Sociéte française de pédiatrie (August 2025): S092–9693, 10.1016/j.arcped.2025.05.010. [DOI] [PubMed] [Google Scholar]
- 18. Legius E., Messiaen L., Wolkenstein P., et al., “Revised Diagnostic Criteria for Neurofibromatosis Type 1 and Legius Syndrome: An International Consensus Recommendation,” Genetics in Medicine 23, no. 8 (2021): 1506–1513, 10.1038/s41436-021-01170-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Santoro C., Picariello S., Palladino F., et al., “Retrospective Multicentric Study on Non‐Optic CNS Tumors in Children and Adolescents With Neurofibromatosis Type 1,” Cancers 12, no. 6 (May 2020): 1426, 10.3390/cancers12061426. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Payne J. M., Pickering T., Porter M., et al., “Longitudinal Assessment of Cognition and T2‐hyperintensities in NF1: An 18‐Year Study,” American Journal of Medical Genetics, Part A 164A, no. 3 (2014): 661–665, 10.1002/ajmg.a.36338. [DOI] [PubMed] [Google Scholar]
- 21. Van Es S., North K. N., McHugh K., and De Silva M., “MRI Findings in Children With Neurofibromatosis Type 1: A Prospective Study,” Pediatric Radiology 26, no. 7 (1996): 478–487, 10.1007/BF01377205. [DOI] [PubMed] [Google Scholar]
- 22. Jandhyala N. R., Garcia M. R., Kim M., Yohay K., and Segal D., “Identifying Lesions of the Corpus Callosum in Patients With Neurofibromatosis Type 1,” Pediatric Neurology 156 (2024): 66–71, 10.1016/j.pediatrneurol.2024.04.009. [DOI] [PubMed] [Google Scholar]
- 23. Taal W., Zick B., Emmer B. J., and van den Bent M. J., “Non‐Optic Glioma‐Like Lesions in Adult Neurofibromatosis Type 1 Patients,” Diagnostics 15, no. 1 (December 2024): 67, 10.3390/diagnostics15010067. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supporting File 1
Supporting File 2
Supporting File 3
Supporting File 4
Supporting File 5
Supporting File 6
Data Availability Statement
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions. Additional supporting information can be found online in the Supporting Information section at the end of this article.