Role of the apparent diffusion coefficient as a predictive factor for tumor recurrence in patients with intracranial epidermoid tumor


 Purpose.

Intracranial epidermoid tumors are slowly growing benign tumors, but due to adjacent critical neurovascular structures, surgical resection is challenging, with the risk of recurrence. The apparent diffusion coefficient (ADC) has been used to evaluate the characteristics of brain tumors, but its utility for intracranial epidermoid tumors has not been specifically explored. This study analyzed the utility of preoperative ADC values in predicting tumor recurrence for patients with intracranial epidermoid tumors.
Methods.

Between 2008 and 2019, 23 patients underwent surgery for intracranial epidermoid tumor, and their preoperative ADC data were analyzed. The patients were divided into two groups: the recurrence group, defined by regrowth of the remnant tumor or newly developed mass after gross total resection on magnetic resonance imaging (MRI); and the stable group, defined by the absence of growth or evidence of tumor on MRI. Receiver operating characteristic (ROC) analysis was used to obtain the ADC cutoff values for predicting tumor recurrence. The prognostic value of the ADC was assessed using Kaplan-Meier curves.
Results.

The minimum ADC values were significantly lower in the recurrence group than in the stable tumor group (P = 0.046). ROC analysis showed that a minimum ADC value lower than 804.5×10− 6 mm2/s could be used to predict higher recurrence risk of intracranial epidermoid tumors. Subtotal resection, younger age, and mean and minimum ADC values lower than the respective cutoffs were negative predictors of recurrence-free survival.
Conclusions.

Minimum ADC values could be useful in predicting the recurrence of intracranial epidermoid tumors.


Introduction
Intracranial epidermoid tumors account for approximately 1% of all brain tumors [1,2]. They are thought to develop from the trapped ectodermal squamous epithelium during neural tube closure, and most of them are generally located off the midline, such as at the cerebellopontine angle (CPA) [1,3,4]. These tumors are composed of keratinized strati ed squamous epithelium lled with keratin debris, lipid, protein, and cholesterol crystals that can spread throughout the cerebrospinal uid space (i.e., cisterns, sulci, and ventricle) [5][6][7]. Although they have a benign nature with slow growth rates, surgical resection is generally conducted if they cause neurologic de cits due to compression of neural structures. However, there are risks of surgical morbidity and recurrence because the tumor capsule adherent to the adjacent critical neurovascular structures makes it di cult to surgically eradicate the tumor [1,2,4,8].
In cases of recurrent symptomatic epidermoid tumors, reoperation is still the mainstay of treatment, and some studies advocate adjuvant radiotherapy due to increased surgical morbidities [8][9][10]. Several studies have investigated factors related to recurrence in intracranial epidermoid tumors [8,11,12], but the understanding of the disease behavior is still limited due to its rarity and pathologic simplicity. The extent of resection (EOR) is the only known prognostic factor predicting tumor recurrence, and there are no reliable factors able to predict tumor behavior before surgery.
The apparent diffusion coe cient (ADC) has been utilized to predict tumor behavior in central nervous system tumors such as glioma [13], meningioma [14], and chordoma [15] because it could re ect tumor cell characteristics. In the management of intracranial epidermoid tumors, preoperative, postoperative, and follow-up magnetic resonance imaging (MRI), including diffusion weighted image (DWI) sequences, are always conducted. These data allow not only detection of the presence of an epidermoid tumor, but also acquisition of ADC values. In this study, we evaluated the behavior of intracranial epidermoid tumors using preoperative ADC values, and determined the ADC cutoff values to predict tumor recurrence.

Study population
We conducted a retrospective single-institution analysis of 27 patients who underwent surgery for newly diagnosed intracranial epidermoid tumors between January 2008 and December 2019. Among them, three patients had extradural epidermoid tumors and one had a frontal convexity epidermoid tumor. Tumors located in the extradural and the frontal convexity area have limited space to spread compared with those located in cisterns or ventricles. As a result, they manifest as a mass with well-circumscribed margins, have low ADC values due to the dense environment, and can be totally removed without recurrence at the surveillance. It is consistent with a previous report that scalp epidermoid tumor has low ADC values than those of intracranial epidermoid tumors [16]. Thus, we classi ed intracranial epidermoid tumors into two categories: the extensive type (i.e., cisterns and ventricles) and the limited type (i.e., extradural and convexity) according to the occupied space, and excluded the latter group from the analysis. The 23 extensive type epidermoid tumors were divided into a stable group and a recurrence group. The stable group was de ned by the absence of growth or evidence of tumor on consecutive MRI follow-ups. The recurrence group was de ned as regrowth of the remnant tumor or newly developed mass after gross total resection based on consecutive MRI follow-ups (Fig. 1). The mean follow-up period was 33.8 months (range 3.6-73 months), and all recurrences except for one case occurred within 2 years. The current study design and the use of clinical data were approved by the Gangnam Severance Hospital institutional review board. All experiments were carried out in accordance with approved guidelines and with the 1964 Helsinki Declaration and its later amendments. The requirement to obtain informed consent was waived, and all data were fully anonymized.
The EOR was based on operative notes and con rmed by the rst postoperative MRI within 3 months of surgery. Gross total resection (GTR) was de ned as complete removal of both capsule and contents; neartotal resection (NTR) was de ned as complete content removal and incomplete capsule removal; subtotal resection (STR) was de ned as incomplete resection of both capsule and contents [17]. Postoperative diffusion-weighted MRI sequences were utilized to con rm the degree of resection. If MRI indicated no recurrence of disease after surgery, imaging studies were typically performed annually.
Two neurosurgeons and two neuroradiologists independently outlined four round or oval regions of interest (ROIs) within the tumor for the evaluation of ADC values ( Fig. 1c and 1i). All continuous sections of the ADC maps that included tumor were evaluated. Each region of interest was positioned carefully to avoid contamination from adjacent tissues. ADC values < 10×10 -6 mm 2 /s were considered as artifacts. A ROI containing lowest ADC value among the four ROIs was chosen for analysis. The mean, minimum, and maximum ADC values were obtained within this same ROI. ADC cutoff values for predicting intracranial epidermoid tumor recurrence were obtained by comparing the preoperative ADC values between the two groups. These cutoff values were then used for recurrence-free survival (RFS) analysis.
The ADC measurements were assessed for interobserver reliability using the interclass correlation coe cient. T1 and T2 weighted image signal intensity and computed tomography (CT) density were evaluated based on comparison with adjacent brain tissue.

Statistical analysis
Mean, maximum and minimum ADC values, patient age at the time of rst surgery, sex, body mass index, EOR, CT density, signal intensity on T1 and T2 weighted images, T1-weighted contrast enhancement, diffusion restriction, and postoperative complications were compared between the two groups using the Student's t-test for continuous variables and Fisher's exact test for nominal variables. The cutoff ADC values were assessed using receiver operating characteristic (ROC) analysis to predict tumor recurrence. Logistic regression analysis was performed to determine independent risk factors for recurrence. Multiple logistic regression analyses were performed on variables with signi cant unadjusted effect on simple logistic regression analysis. RFS was analyzed using Kaplan-Meier curves and compared between the groups using log-rank tests with the following variables: mean and minimum ADC cutoff values, age, and EOR. All statistical analyses were performed using IBM SPSS statistics version 25.0 (IBM Corp, Armonk, NY, USA). Two-tailed P-values <0.05 were considered statistically signi cant.
Twenty cases involved the CPA area, and three cases were in the 4th ventricle or con ned to the ambient cistern. Most operations were via lateral suboccipital craniotomy (34.8%) or posterior petrosal approach (34.8%).
There were 12 cases in the stable group (52%) and 11 cases (48%) in the recurrence group. Most STR cases recurred, and all recurrence events occurred within 2 years except for one patient (mean 12.1 months, range 5-24.5 months). Patient characteristics are summarized in Table 1. The mean age was lower in the recurrence group than in the stable group (P = 0.024). The extent of resection was analyzed into two different ways. When considering GTR, NTR, and STR separately, the rate of STR was higher in the recurrence group than in the stable group (P = 0.021). Likewise, when NTR and STR were considered together as cases of "non-total resection", the rate of non-total resections was higher in the recurrence group than in the stable group (P = 0.036). The minimum ADC values were signi cantly lower in the recurrence group than in the stable group (P = 0.046). The interclass correlation coe cient for interobserver reliability of the minimum ADC values was 0.881 (95% CI, 0.720-0.950). There were no signi cant differences in sex, BMI, postoperative complication, CT density, T1, T2 signal intensity, T1 weighted contrast enhancement, diffusion restriction, or mean and maximum ADC values between the two groups. Most intracranial epidermoid tumors showed as hypodense on CT, hypointense on T1, hyperintense on T2 MRI sequence, non-enhancing, and restricted on DWI.
These cutoff ADC values and the parameters found to be signi cantly different in the previous analysis (age and EOR) were used in logistic regression analysis. Logistic regression analysis with adjustment showed that a minimum ADC value ≤804.5×10 -6 mm 2 /s was independently associated with recurrence of intracranial epidermoid tumors (P = 0.039; Table 2). Age less than 40 years, EOR, and mean ADC value ≤1043×10 -6 mm 2 /s were associated to P-values of 0.027, 0.027, and 0.029, respectively, in simple logistic regression, but did not reach statistical signi cance after adjustment.

Prediction of recurrence-free survival in intracranial epidermoid tumors
The predictive factors associated with recurrence of intracranial epidermoid tumors in Table 2 were used in a Kaplan-Meier survival analysis of recurrence. The log-rank test showed that mean ADC value ≤1043×10 -6 mm 2 /s (P = 0.029), minimum ADC value ≤804.5×10 -6 mm 2 /s (P = 0.018), EOR (STR) (P = 0.046), and age less than 40 years were signi cantly associated with poor RFS (Fig. 3).

Discussion
To the best of our knowledge, this is the rst study to show the utility of ADC values for predicting intracranial epidermoid tumor recurrence. Epidermoid tumors are slow-growing tumors often involving the posterior fossa, such as the CPA, and cause neurologic de cits when leading to mass effects or stretching onto adjacent neurovascular structures [5,8,12]. Although there are no established treatment guidelines, maximal safe surgical resection is generally conducted, and reoperation is often considered for recurrent cases. Due to its benign nature and pathologic simplicity, many studies related to intracranial epidermoid tumors focus on surgical techniques, cranial nerve function preservation, and recurrence after surgery [2,5,8,12,[18][19][20][21]. The EOR was shown to be a signi cant prognostic factor for recurrence in these studies, but there are no known preoperative factors that can predict tumor behavior. ADC values have been used to represent tumor characteristics in several central nervous system tumors [15,22,23], and DWI sequence is always taken for detection of epidermoid tumor recurrence [24]. Thus, in the present study, we analyzed 23 patients using ADC values, hypothesizing that they could be useful in predicting the recurrence of intracranial epidermoid tumors.
There have been many debates about the optimum extent of resection in intracranial epidermoid tumors [2,5,8,25]. The proliferation of strati ed squamous epithelium is considered to be the cause of tumor recurrence, leading to accumulation of acellular debris such as keratin, protein, and lipid [1,8,9]. Considering the pathogenesis of this tumor, many studies have revealed that complete excision, including of the tumor capsule, reduces the risk of recurrence compared with STR [2,5,8]. Our case series con rms that patients treated with total resection showed lower recurrence rates than those treated with non-total resection (P = 0.036). While many neurosurgeons attempt GTR to minimize the recurrence rate, surgical morbidity must also be considered. The capsule can be highly adherent to adjacent critical neurovascular structures, and excessive surgical resection can lead to serious complications such as cranial nerve palsy and infarction [2,8,18]. Thus, some authors support intentional STR if an attempt at GTR seems too risky, especially in older patients with signi cant medical comorbidities [8,20].
A recent meta-analysis reported that the recurrence rate of intracranial epidermoid tumors is roughly one in ten patients, and some studies showed no recurrence even after subtotal resection [8]. However, several studies included in that meta-analysis obtained follow-up imaging only when symptoms occurred and did not perform MRI annually, which might make their interpretation problematic [2,19,26]. In the current study, we annually repeated MRI studies to radiologically screen for tumor recurrence, and we found that most STR cases do recur, most recurrences occurring within 2 years of surgery. Our results also show that younger age is signi cantly correlated with higher probability of tumor recurrence. Although epidermoid tumors generally show benign behavior when GTR is achieved, it should be emphasized that they can often recur after STR, as our study shows. Therefore, the optimum EOR should be tailored for each patient considering recurrence risk, age, comorbidities, and surgical morbidities. For instance, most patients, in particular younger ones, should receive GTR to reduce the recurrence rate, but STR might be chosen to avoid perceived morbidities while being aware of the recurrence risk.
Some remnant epidermoid tumors have a benign course without recurrence, but some do recur even after GTR [8]. The understanding of the natural course of this tumor is still limited and a grading system to distinguish its characteristics is not even available, due to its pathological simplicity. It is generally thought to grow slowly and to be benign, but some cases recur rapidly and rarely undergo malignant transformation [27]. Several studies have also reported that 'white epidermoid tumors' have more protein components within the capsules than others [6,28]. That is, there are differences in the characteristics of epidermoid tumors, such as in the production and accumulation of their components, although they are all classi ed as "intracranial epidermoid tumors". To re ect these differences, we measured the ADC values using preoperative DWI because the ADC has been shown to represent the nature of the tumor in several brain tumors [13][14][15]. Diffusion measurements re ect intra-and extra-cellular water motion and could indicate tumor characteristics [29]. We found that the mean ADC values ranged roughly from 750 to 1500 (10 -6 mm 2 /s), and the minimum ADC values ranged from 600 to 1200 (10 -6 mm 2 /s), which is consistent with previous epidermoid ADC studies [16,30,31]. We focused on the minimum ADC values because the tumor ROIs often contain cerebrospinal uid, which could increase the maximum and mean ADC values. We also distinguished intracranial extradural epidermoid tumors and convexity tumors from cisternal and ventricular tumors, which tend to spread into the subarachnoid space, and classi ed them into the "limited" and "extensive" types, respectively. Limited type epidermoid tumors have low ADC values with well circumscribed margins in a limited space, and were totally removed without di culty in our cases, again consistently with previous reports [16,17]. However, extensive type epidermoid tumors, such as those located in the CPA area, have variable ADC values and prognosis.
Using these ADC values, we analyzed whether they re ected tumor behavior and could be predictive factors in intracranial epidermoid tumors of the extensive type. Our results show that the minimum ADC values in the recurrence group were signi cantly lower than in the stable group. Moreover, cases with minimum ADC values lower than the cutoff showed signi cantly shorter RFS (mean 19.1 months) than those with higher values (mean 50.6 months). That is, we could predict the prognosis of intracranial epidermoid tumors using their preoperative ADC maps. These ndings are similar to those of previous reports showing that aggressive clival chordoma and high-grade meningioma and glioma have lower ADC values than those of less aggressive or low-grade tumors [13][14][15]. Therefore, close MRI follow-up is recommended for cases with minimum ADC values lower than the cutoff to screen for recurrence after surgery.
There are several limitations to this study. First, its design was retrospective, with a small sample size. Intracranial epidermoid tumor is a rare disease, making large-scale prospective studies di cult. Second, the follow-up period was not long enough, considering the benign behavior of intracranial epidermoid tumors. Although all cases of recurrence, except for one, occurred within 2 years, a further long-term follow-up investigation is required in the future. Third, most tumors were located in the CPA area rather than in the parasellar region. Fourth, there were no cases of malignant transformation, and we could not analyze their ADC values. Despite these limitations, our results suggest that preoperative ADC values could be useful in predicting the recurrence of intracranial epidermoid tumors. In this respect, our study is the rst to analyze the signi cance of ADC values as recurrence predictors in intracranial epidermoid tumors.
In conclusion, we found that minimum ADC values and age were signi cantly lower in the recurrence group than in the stable group. Minimum ADC values lower than the cutoff value of 804.5×10 -6 mm 2 /s and patients younger than 40 years showed signi cantly poorer RFS than those with higher ADC values and older than 40 years. Thus, preoperative ADC values can be useful in predicting the recurrence of intracranial epidermoid tumor as well as in diagnosis.   Figure 1 a A 54-year-old woman was diagnosed with an intracranial epidermoid tumor and classi ed in the stable group. Preoperative T2-weighted imaging showed a tumor located in the right CPA area and spread into the ambient cistern. b Preoperative diffusion weighted imaging showed diffusion restriction in the same area. c The ROI outlined in yellow on the ADC map represents decreased water diffusivity (mean ADC 1067 × 10-6 mm2/s, minimum ADC 997 × 10-6 mm2/s). d Postoperative diffusion weighted imaging shows remnant tumor at the right ambient cistern. Surgery was performed via lateral suboccipital craniotomy, and subtotal resection was achieved. e Diffusion weighted imaging obtained a year after surgery showed stable disease. f Diffusion weighted imaging 5 years after surgery still indicates stable status. g A 35-year-old man diagnosed with an intracranial epidermoid tumor and classi ed in the recurrence group. Preoperative T2-weighted imaging showed a tumor located from the suprasellar cistern to the right CPA area. h Preoperative diffusion weighted imaging showed diffusion restriction in the same area. i The ROI outlined in yellow on the ADC map represents decreased water diffusivity (mean ADC 977 × 10-6 mm2/s, minimum ADC 776 × 10-6 mm2/s). j Postoperative diffusion weighted imaging shows remnant tumor at the suprasellar cistern. Surgery was performed via the posterior petrosal approach and subtotal resection was achieved. k Diffusion weighted imaging obtained a year after surgery showed an increase in size of the remnant tumor. l Diffusion weighted imaging 4 years after surgery reveals progressive recurrence. CPA, cerebellopontine angle; ADC, apparent diffusion coe cient; ROI, region of interest Figure 2 ROC curves for the ADC values differentiating the recurrence group from the stable group. ROC, receiver operating characteristic; ADC, apparent diffusion coe cient