Clivus length distinguishes between asymptomatic healthy controls and symptomatic adult women with Chiari malformation type I

Abstract

Purpose

While the presence of cerebellar tonsillar descent in radiological images has been used as evidence of Chiari malformation type I (CMI), tonsillar ectopia alone is insufficient to identify individuals with symptomatic CMI. This study sought to identify differences in brain morphology between symptomatic CMI and healthy controls in adult females.

Methods

Two hundred and ten adult females with symptomatic CMI and 90 age- and body mass index-matched asymptomatic female controls were compared using seven brain morphometric measures visible on magnetic resonance images. The CMI and control groups were divided into four subgroups based on the tonsillar position (TP) relative to the foramen magnum: group 1 was made up of healthy controls with normal TP (TP < 0 mm); group 2 was comprised of control individuals with low-lying TP (1–5 mm); group 3 was comprised of symptomatic CMI patients with low-lying TP (1–5 mm); group 4 contained symptomatic CMI patients with severe tonsillar descent (6–13 mm).

Results

All morphometrics for symptomatic CMI with severe tonsillar descent were significantly different than those for both control groups. The CMI group with low-lying TP was significantly different for four measures when compared to controls with normal TP. However, only clivus length was statistically different between the CMI and healthy control groups with low-lying TP.

Conclusion

This study demonstrates that clivus length distinguishes adult female healthy individuals with low-lying tonsils from those with symptomatic CMI. Further investigation is required to understand the importance of a shorter clivus length on CMI symptomatology and pathophysiology.

Introduction

Chiari malformation was first described in 1891 when Hans Chiari associated a malformation of the brainstem and cerebellum with congenital hydrocephalus [1]. Since then, four distinct types of Chiari malformation have been identified (types I through IV [2]). The current study focuses on type I morphology. Chiari malformation type I (CMI) is an anatomical anomaly in which the cerebellar tonsils are located 3–5 mm below the foramen magnum (FM; see Fig. 1 [3]). CMI patients present with a host of symptoms. Commonly reported symptoms include occipital headache, neck pain, cognitive dysfunction, and balance problems [3–5]. The prevalence of CMI in the USA has been estimated at approximately 0.08%, but this proportion changes when different diagnostic standards are used [6–8].

Fig. 1.

Mid-sagittal view of the cerebellum (outlined in blue) for a CMI individual showing tonsillar descent below the FM (left) and healthy control (right). The red line indicates the McRae line, and the yellow arrows show the position of the tonsils relative to the McRae line

While the presence of descent of the tonsils in radiological images has been used as evidence of CMI, previous studies show that the position of the cerebellar tonsils alone is insufficient to identify individuals with symptomatic CMI [6, 8]. Due to the weak association between tonsillar position and symptom severity in CMI [9], several research groups have attempted to identify additional morphological markers that are unique to individuals suffering from CMI [10–18]. For example, Houston et al. reported several morphometric brain parameters to be significantly different between CMI and healthy controls [14]. Also, Biswas et al. reported greater crowding of the posterior cranial fossa (PCF) in CMI patients when compared to controls [19].

Smith et al. reported that approximately 1–2% of individuals with tonsillar position (TP) greater than 5 mm remain asymptomatic [6]. Furthermore, Strahle et al. found that there are 10 times more asymptomatic individuals with TP greater than 5 mm than symptomatic individuals [8]. Moreover, Milhorat et al., Allen et al., and Houston et al. reported that TP was poorly correlated with the symptom severity of individuals with characteristics of CMI [3, 9, 20]. In contrast, Kavak et al. determined the impact of morphological measurements on CMI symptoms and found that the amount of tonsillar herniation was positively correlated with the modified Asgari score [21]. Huang et al. determined the utility of TP to predict cough associated headaches in CMI individuals. They found that a much higher extent of TP (TP ≥ 14 mm) was required to be predictive of a cough-associated headache in CMI patients [22]. Khalsa et al. analyzed morphological data from 51 symptomatic and 51 asymptomatic Chiari malformation patients that were matched for age at diagnosis, sex, TP, and tonsil morphology. They found no significant differences between symptomatic and asymptomatic Chiari patients in terms of either two-dimensional (2D) or three-dimensional (3D) morphometric measures [23]. In 2019, Bolognese et al. collected opinions via a questionnaire from a panel of 63 recognized international experts with a collective surgical experience of over 15,000 CMI cases [24]. The 90 questions focused on a variety of CMI topics including pathophysiology, epidemiology, signs/symptoms, and comorbidities. More than 85% of the experts responded positively to reject the 5 mm rule used as the basis for diagnosing CMI [24]. This finding raises more questions about the accuracy of the 5 mm criterion in terms of the definition, diagnostic criteria, classification, and treatment of CMI [24]. Overall, there is weak evidence that tonsillar position alone is indicative of either the presence or severity of CMI related symptoms. Thus, a study comparing symptomatic CMI subjects and healthy controls is warranted to determine the morphometric differences between those that do and do not satisfy the 5 mm criterion.

The goal of the present study was to determine if there are morphological differences beyond TP between incidentals and symptomatic CMI patients. At present, clinicians do not have a reliable morphometric measurement to predict the symptom severity of CMI. Discovery of morphometric differences between incidental TP and symptomatic CMI may help elucidate which morphometric features are indicative of symptoms specific to CMI.

Methods

In the present study, healthy individuals and those with CMI were evaluated. These groups were divided into four subgroups based on the position of the tonsils relative to the FM: the first group was made up of healthy controls with normal TP (controls with TP between − 3 and − 11 mm: TP < 0 mm), the second was control individuals with low-lying TP (TP between 1 and 5 mm), the third group was symptomatic CMI with low-lying TP (TP between 1 and 5 mm), and the fourth group was symptomatic CMI with severe tonsillar descent (TP between 6 and 13 mm). We compared seven brain morphometric measures in the various subgroups to identify measures beyond TP that may be useful for CMI diagnosis and clinical evaluation.

Participants

This study included 210 adult females diagnosed with CMI (mean age = 37 ± 10 years, mean body mass index (BMI) = 32 ± 8) and 90 non-Chiari healthy female controls (mean age = 34 ± 12 years, mean BMI = 28 ± 6). The images employed in this study were previously reported [12, 14, 19]. All symptomatic CMI patient data—including demographics and health-related data—were voluntarily supplied by patients through Chiari1000, a Web-accessible database developed at The University of Akron in which diagnosis, symptomatology, cognitive functions, and socioeconomic characteristics, as well as radiologic images of CMI subjects, were stored. De-identified MR images were submitted through online sharing, direct mail, or a signed release that enabled images to be acquired from various hospitals. Images and demographic information for control subjects were obtained from several sources, including members of the WU/Minn Project of the Human Connectome Project consortium (Washington University in St. Louis, University of Minnesota, and Oxford University), along with Akron General Medical Center (Akron, Ohio, [25]). All healthy control subjects had no known history of conditions that may affect brain morphology and were free of CMI diagnosis. Due to the limited number of image datasets for male and juvenile subjects in the patient database, adult males and children were excluded from this study.

Four groups were created from the available image sets based on the position of the cerebellar tonsil (as shown in Table 1). Comparisons were made between these groups to identify any similarities/differences.

Table 1.

Description of groups included in this study

Status	Mean age, mean BMI	Group ID (no. of samples)	Tonsillar position	Mean age, mean BMI	Subset	No. of samples
CMI subjects (n = 210)	37 ± 10 years, mean BMI = 32 ± 8	CMI > 5 (n = 127)	Severe tonsillar descent (6–13 mm)	36 ± 10 years, mean BMI = 30 ± 8	Validation set	63
					Holdout set	64
		CMI < 5 (n = 83)	Low-lying TP (1–5 mm)	39 ± 10 years, mean BMI = 29 ± 7	Validation set	41
					Holdout set	42
Healthy controls (n = 90)	34 ± 12 years, mean BMI = 28 ± 6	CON+TP (n = 41)	Low-lying TP (1–5 mm)	35 ± 11 years, mean BMI = 28 ± 5	n/a*	n/a*
		CON-TP (n = 49)	Normal TP (< 0 mm)	34 ± 12 years, mean BMI = 28 ± 6	n/a*	n/a*

^*Not applicable. No validation and holdout sample sets were created for the healthy control groups because of their small sample sizes

Measurement software and image selection

All measurements were performed using MorphPro [12] and CerePro [19], custom in-house software developed in MATLAB (MathWorks, Natick, MA) at the University of Akron. Measurements were performed on mid-sagittal T1-weighted (293) and T2-weighted (7) MR images acquired using scanners manufactured by Philips, Siemens, Toshiba, and General Electric. Mid-sagittal images were selected using 3D multi-planar reconstruction in OsiriX (Pixmeo SARL, Switzerland). The criterion used to determine the mid-sagittal plane of the image was based on the visibility of at least three of the following four structures: the splenium of the corpus callosum, the pituitary, the cerebral aqueduct, and the genu of the corpus callosum [12]. Due to poor image resolution or lack of sagittal images, we were unable to perform morphometric measurements on a total of 144 image sets. Thus, these images were excluded from the analysis and we do not expect this to be a source of selection bias. Although without a measurement of TP, we cannot group the 144 subjects by TP, and thus, we cannot confirm the lack of selection bias.

Morphometric measurement

Seven 2D brain morphometric parameters (Fig. 2) were evaluated for each subgroup:

1. Tonsillar position—the perpendicular distance between the McRae line and the most inferior portion of the tonsils.

2. PCF area—the area bounded by the tentorium, the occipital bone, the McRae line, and the clivus.

3. Clivus length—distance from the dorsum sellae to the basion.

4. Fastigium height—the perpendicular distance between the McRae line and the fastigium of the cerebellum.

5. Boogard angle—the angle between the clivus and the McRae line.

6. Anterior cerebrospinal fluid (CSF) area—the CSF space in the upper cervical spinal canal anterior to the spinal cord between the FM and the inferior limit of the C2 vertebrae

7. Posterior CSF area—the CSF space in the upper cervical spinal canal posterior to the spinal cord between the FM and the inferior limit of the C2 vertebra.

Fig. 2.

Morphometric measurements obtained from a mid-sagittal T1-weighted MRI image: (1) tonsillar position, (2) PCF area, (3) clivus length, (4) fastigium height, (5) Boogard angle, (6) anterior CSF area, and (7) posterior CSF area

Five different trained morphometric operators performed the measurements due to a large number of images. For inter-rater reliability tests, the operators were blinded to the CMI diagnosis of the subject and randomly assigned thirty images for analysis. For intra-rater consistency, the operators were provided with 15 random cases, where those 15 cases were measured 3 times, with a minimum of 3 days in between measurements. Two-way mixed intraclass correlation coefficients were computed for absolute agreement and the average intraclass correlation coefficients value ranged from 0.72 to 0.97. Thus, there was a good agreement between the raters and between measurements by the same rater.

Self-reported clinical parameters

Additional data from Chiari1000 in the form of self-reported symptoms, surgical history, prevalent conditions, and selected neuropsychological scales for the CMI groups with TP > 5 mm and TP < 5 mm were compared (see Table 2 for all self-reported clinical parameters).

Table 2.

Self-reported clinical parameters

(A) Self-reported symptom severity before surgery (rated from “1” to “5”, where “1” indicates no symptoms, “2” mild symptoms, “3” moderate symptoms, “4” severe symptoms, and “5” very severe symptoms)
(B) Surgical outcome (rated from “1” to “5”, where “5” indicates completely resolved, “4” improved significantly, “3” improved slightly, “2” no change, and “1” got worse)
(C) Hypermobility score	(J) Cervical instability	(P) Double vision
(D) Neck pain	(K) Seizures	(Q) Sensitivity to light
(E) Decompression surgery	(L) Upper back pain	(R) Depression
(F) Syringomyelia	(M) Numbness	(S) Anxiety
(G) Ehlers–Danlos syndrome (EDS)	(N) Weakness	(T) Stress
(H) Pseudotumor cerebri (PTC)	(L) Dizziness	(U) Loneliness
(I) Migraine pain	(O) Blurred vision
(V and W) Immediate and delayed recall memory as assessed using the Rey Auditory Verbal Learning Task (RAVLT)
(X) The total score index of the Short Form–McGill Pain Questionnaire–Revised (a comprehensive, standardized self-report pain index)

Statistical analysis

Morphometric measurements for the different groups were compared using Student’s t test. Student’s t test was used as all morphometric parameters in the four groups (CON−TP, CON+TP, CMI < 5, and CMI > 5) were normally distributed. Due to differences in the sample size, the symptomatic CMI groups were divided into validation and holdout samples to confirm whether the results obtained for our validation sample generalized to our holdout sample (see Table 1). Individuals in groups CMI > 5 and CMI < 5 were randomly sorted into validation and holdout sets, while the control groups were not divided because of the smaller group sample sizes. Two separate comparisons were made. The first comparison used the CMI validation set: CON−TP vs. CON+TP, CON−TP vs. CMI < 5, CON−TP vs. CMI > 5, CON+TP vs. CMI < 5, CON+TP vs. CMI > 5, and CMI < 5 vs. CMI > 5 (see Table 1 for details of the groups). The second comparison used the CMI holdout set.

The self-report and scales data for groups CMI > 5 and CMI < 5 were compared. The chi-square test was used to determine differences in categorical data (decompression surgery, syringomyelia, EDS, PTC, migraine pain, cervical instability, seizures, upper back pain, numbness, weakness, dizziness, blurred vision, double vision, and sensitivity to light). Student’s t tests were used to identify differences in quantitative data (symptom severity, surgical outcome, hypermobility scores, neck pain, depression, anxiety, stress, loneliness, immediate memory, delayed memory, and McGill pain score). Correlations were performed to determine the strength of associations between the seven morphometric parameters and the 25 self-reported clinical parameters (SRCPs). Separate correlations were performed for the two CMI groups. Additionally, collinearity assessments were performed on each morphometric parameter to determine whether or not the measurements were inter-related. The variance inflation factor ranged from 1.1 to 2.7, signifying a low to moderate level of multicollinearity between the morphometric measurements. Statistical significance was assumed at p < 0.05.

Results

Morphometric group differences

Presented values of groups are expressed as mean ± standard deviation. The mean age and BMI of CMI subjects were 37 ± 10 years and BMI = 32 ± 8, respectively, while the mean age and BMI of the healthy controls were 34 ± 12 years and BMI = 28 ± 6, respectively. The mean TP for CMI > 5 (range of TP = 6–13 mm) and CMI < 5 (range of TP = 1–5 mm) for the validation set was 8.8 ± 2.1 and 3.9 ± 1.0 mm, respectively, while for the holdout set, the mean TP for CMI > 5 and CMI < 5 was 9.1 ± 2.0 and 3.2 ± 1.3 mm, respectively. For the CON−TP (range of TP = −3 to − 11 mm) and CON+TP (range of TP = 1–5 mm) groups, the mean TP was − 5.5 ± 1.7 and 1.5 ± 1.1 mm, respectively.

For comparisons involving individuals in groups CMI > 5 and CMI < 5, a morphometric parameter was considered significant if and only if it survived the holdout set analysis. All morphometric parameters were found to be significantly different when comparing CON−TP vs. CMI > 5 and CON+TP vs. CMI > 5 in both the validation and holdout sets (see Table 3). The PCF area was shown to be significantly smaller in the group with CMI > 5 (3167 ± 190 mm²) as compared to the CON−TP (3421 ± 318 mm², p = 9e⁻⁷) and CON+TP (3293 ± 249 mm², p = 0.01) groups. The clivus length, fastigium height, anterior CSF area, and posterior CSF area were also shown to be significantly smaller in the CMI > 5 group as compared to the CON−TP and CON+TP groups (see Table 3 for details). The Boogard angle, in contrast, was found to be significantly larger in the CMI > 5 (125.7 ± 9.0°) group as compared to groups CON−TP (120.9 ± 5.9°, p = 0.01) and CON+TP (119.5 ± 5.7°, p = 0.0007).

Table 3.

Descriptive statistics for the different group comparisons

Validation set (CON-TP [n = 49], CON+TP [n = 41], CMI < 5 [n = 41], CMI > 5 [n = 63])
Measurement	CON-TP	CON+TP	CMI < 5	CMI > 5	p (CON-TP vs CON+TP)	p (CON-TP vs CMI < 5)	p (CON-TP vs CMI > 5)	p (CON+TP vs CMI < 5)	p (CON+TP vs CMI > 5)	p (CMI < 5 vs CMI > 5)
PCF area (mm2)	3421(318)	3293(249)	3175(256)	3167(190)	0.04*	0.0001*	9e−7*	0.04	0.01*	0.86
Clivus length (mm)	42.2(4.3)	42.5(3.9)	39.8(4.0)	38.0(4.0)	0.79	0.006*	7e−7*	0.002*	2e−7*	0.03*
Fastigium height (mm)	31.2(2.3)	28.8(1.6)	28.6(2.4)	26.2(2.2)	7e−8*	1e−6*	1e−20*	0.74	6e−10*	1e−6*
Boogard Angle (°)	120.9(5.9)	119.5(5.7)	121.2(7.2)	125.7(9.0)	0.24	0.83	0.01*	0.24	0.0007*	0.03
Anterior CSF area (mm2)	101(23)	92(22)	79(25)	81(22)	0.07	0.0002*	3e−5*	0.03	0.02*	0.79
Posterior CSF area (mm2)	116(36)	117(43)	103(35)	84(35)	0.84	0.13	3e−5*	0.13	0.0002*	0.02*
Holdout set (CMI < 5 [n = 42], CMI > 5 [n = 64])
Measurement	–	–	CMI <5	CMI >5	–	p (CON-TP vs CMI < 5)	p (CON-TP vs CMI >5)	p (CON+TP vs CMI < 5)	p (CON+TP vs CMI > 5)	p (CMI < 5 vs CMI > 5)
PCF area (mm2)	–	–	3265(270)	3168(309)	–	0.01*	5e−5*	0.62	0.02*	0.09
Clivus length (mm)	–	–	40.4(3.2)	38.7(4.3)	–	0.02*	3e−5*	0.01*	8e−6*	0.03*
Fastigium height (mm)	–	–	28.8(2.1)	26.3(2.3)	–	1e−6*	9e−20*	0.91	3e−8*	7e−8*
Boogard Angle (°)	–	–	122.4(7.7)	124.3(8.0)	–	0.30	0.03*	0.05	0.003*	0.31
Anterior CSF area (mm2)	–	–	88(23)	79(19)	–	0.01*	9e−7*	0.25	0.003*	0.04
Posterior CSF area (mm2)	–	–	109(34)	79(23)	–	0.41	4e−8*	0.37	8e−7*	4e−6*

^*Represents a statistically significant difference and numbers in column represents mean (standard deviation) of the measurement

For the comparison between CON−TP and CMI < 5, the PCF area (CMI < 5: 3175 ± 256 mm² vs CON−TP: 3421 ± 318 mm², p = 0.0001), clivus length (CMI < 5: 39.8 ± 4.0 mm vs CON−TP: 42.2 ± 4.3 mm, p = 0.006), fastigium height (CMI < 5: 28.6 ± 2.4 mm vs CON−TP: 31.2 ± 2.3 mm, p = 1e⁻⁶), and anterior CSF area (CMI < 5: 79 ± 25 mm² vs CON−TP: 101 ± 23 mm², p = 0.0002) were found to be significantly smaller in the CMI < 5 group as compared to group CON−TP (see Table 3). However, no significant difference was found between CON−TP and CMI < 5 for Boogard angle and posterior CSF area (p > 0.05).

The comparison CON+TP vs. CMI < 5 revealed only the clivus length (CMI < 5: 39.8 ± 4.0 mm vs CON+TP: 42.5 ± 3.9 mm, p = 0.002) to be significantly different between the two groups. Other morphometric parameters (PCF area, fastigium height, Boogard angle, anterior CSF area, and posterior CSF area) did not survive the holdout analysis.

Comparison of morphometric parameters between CMI individuals in CMI > 5 and CMI < 5 revealed the clivus length (CMI > 5: 38.0 ± 4.0 mm vs CMI < 5: 39.8 ± 4.0 mm, p = 0.03), fastigium height (CMI > 5: 26.2 ± 2.2 mm vs CMI < 5: 28.6 ± 2.4 mm, p = 1e⁻⁶) and posterior CSF area (CMI > 5: 84 ± 35 mm² vs CMI < 5: 103 ± 35 mm², p = 0.02) to be significantly smaller in the CMI > 5 group as compared to group CMI < 5. The PCF area, the Boogard angle, and the anterior CSF area showed no significant statistical differences between the two groups (p > 0.05, see Table 3).

A comparison between the control groups showed the PCF area (CON−TP: 3421 ± 318 mm² vs CON+TP: 3293 ± 249 mm², p = 0.04) and fastigium height (CON−TP: 31.2 ± 2.3 mm vs CON+TP: 28.8 ± 1.6 mm, p = 7e⁻⁸) to be significantly smaller in the CON+TP group as compared to the CON−TP group (see Table 3). The clivus length, Boogard angle, and the anterior and posterior CSF area were not found to be significantly different (p > 0.05).

Self-reported clinical parameters

There were no non-Chiari pathology cases in the database. All reported symptoms in the Chiari1000 database were Chiari-related. The SRCPs for CMI subjects in groups CMI < 5 and CMI > 5 were compared. No significant difference was found between the two groups for all the SRCPs. Additionally, no statistically significant correlation was observed between TP and the 25 SRCPs for either CMI < 5 or CMI > 5.

For the CMI < 5 group, anxiety was the only measure that was significantly correlated with anterior CSF area (r = − 0.37, p = 0.03). No significant correlation was found between morphometrics and the other SRCPs. Nonetheless, some SRCPs demonstrated a trend with morphometrics, though the correlation did not survive the holdout set analysis. For example, a trend was found between McGill pain and the fastigium height (r = − 0.36, p > 0.05), clivus length (r = − 0.37, p > 0.05), and PCF area (r = − 0.37, p > 0.05). Additionally, neck pain demonstrated a trend with clivus length (r = − 0.42, p > 0.05).

For the group with CMI > 5, clivus length was the only morphometric parameter to be significantly correlated with weakness (r = − 0.33, p = 0.009) and hypermobility score (r = − 0.36, p = 0.004). No other correlations survived the holdout set analysis, though the clivus length further showed a trend with double vision (r = − 0.30, p > 0.05), migraine pain (r = − 0.30, p > 0.05), depression (r = − 0.50, p > 0.05), anxiety (r = − 0.50, p > 0.05), stress (r = − 0.51, p > 0.05), and McGill pain (r = − 0.30, p > 0.05).

Discussion

This study identified morphometric differences between individuals with varying TP. As expected, all morphometrics for CMI subjects with greater tonsillar descent (CMI > 5) were significantly different than those for the two control groups (CON−TP and CON+TP). In addition, four morphometric measures significantly differed between CMI subjects with low-lying TP and control subjects having a smaller or negative TP (CON−TP). However, clivus length was the only morphometric parameter to differentiate CMI < 5 from controls with low-lying tonsils (CON+TP). In addition, we found two significant morphological differences between the two control groups having different TPs (PCF area and fastigium height) and three significant morphometric differences between groups CMI < 5 and CMI > 5 (clivus length, fastigium height, and posterior CSF area). While our data identified morphological differences between the two CMI groups, comparisons of SRCPs between the two groups revealed no statistically significant differences. This finding suggests that in this cohort of CMI patients, the extent of the descent of the cerebellar tonsils has a minimal impact on reported symptoms, symptom severity, the risk for surgery, surgical outcome, psychological impact, or presence of related conditions. Despite this, we found significant correlations between other morphometric measures and the clinical parameters (for example, clivus length correlated with weakness).

Morphometric findings

The PCF has been extensively studied in CMI subjects, where several studies report a smaller PCF area in CMI patients [3, 10, 14, 18, 19, 26–30]. In the present study, the PCF area was found to be significantly different between the two healthy control groups (CON−TP vs. CON+TP) as well as in healthy controls vs. symptomatic CMI patients (CON−TP vs. CMI < 5, CON−TP vs. CMI > 5, and CON+TP vs. CMI > 5). However, no significant difference was found for the PCF area for two other comparisons: CON+TP vs. CMI < 5 and CMI < 5 vs. CMI > 5. Our results suggest that the PCF area is reduced as the length of the tonsils increases, with the smallest PCF area found in the group with the largest tonsillar descent (CMI > 5, see Fig. 3a). Reduction in the clivus length has been thought to be associated with a reduction in the PCF area [13]. Fernandes et al. reported that a 15-mm reduction in the clivus length reduces the PCF area by about 1.7% [13]. This result is corroborated by our findings, as we found a positive correlation between the clivus length and the PCF area (r = 0.49, p = 1e⁻¹⁹, see Fig. 4a).

Fig. 3.

Bar plots showing the distribution of the four different groups for a PCF area, b clivus length, c fastigium height, and d anterior and posterior CSF area

Fig. 4.

Correlation between clivus length and a PCF area, b fastigium height, c anterior CSF area, and d posterior CSF area. The correlation was performed by combining all the CMI and control groups

The clivus, which develops from the fusion of the basioccipital and basisphenoid bone at the sphenooccipital synchondrosis [31], has been a bone of interest to CMI researchers since Milhorat et al. reported a shorter clivus in individuals with CMI [3]. The findings ofMilhorat et al. were confirmed by subsequent investigations [10, 11, 13–18, 28, 32]. In the current study, we found clivus length to be significantly different in all group comparisons except for the control group comparison (CON−TP vs. CON+TP). Interestingly, for the comparison of the groups with low-lying TP (CON+TP vs. CMI < 5), the clivus was the only morphometric parameter that survived the holdout set analysis. This finding is significant, as it suggests that clivus length could be used as a morphometric parameter to distinguish between symptomatic CMI and incidentals with low-lying tonsils (i.e., in circumstances where the tonsillar position is limited as a predictor of diagnosis). Another important finding from our analysis is that the clivus length was similar in the comparison between CON−TP and CON+TP, even though the difference in tonsillar length between the two groups was ~ 7 mm (− 5.5 mm for CON−TP, and 1.5 mm for CON+TP). This further suggests that a reduction in the tonsillar length may not be directly related to a reduction in clivus length for the control groups. However, for symptomatic Chiari patients, the clivus length tends to become shorter as the tonsillar length increases (see Fig. 3b). A recent evolutional hypothesis proposes that CMI is primarily caused by the abnormal development of segments of the clivus [33]. Understanding the developmental changes of the clivus that occur in symptomatic CMI individuals may help elucidate the etiology of CMI. To further understand the relationship between the clivus length and other morphometric parameters, we correlated clivus length with PCF area, fastigium height, anterior CSF area, and posterior CSF area. We found that the clivus was moderately positively correlated with PCF area (r = 0.49, p = 1e⁻¹⁹) and fastigium height (r = 0.41, p = 9e⁻¹⁴) but was weakly positively correlated with anterior CSF space (r = 0.18, p = 0.004) and posterior CSF space (r = 0.32, p = 2e⁻⁷). Thus, individuals having a shorter clivus (as reported in symptomatic CMI individuals) are likely to also have a smaller PCF area, fastigium height, anterior, and posterior CSF areas (see Fig. 4).

In this study, the fastigium height was shown to decrease with an increase in tonsil length (see Fig. 3c). This trend is further supported by the finding that the fastigium height was similar in the two groups with low-lying tonsils (CON+TP vs. CMI < 5). This finding suggests that the position of the cerebellum is lower relative to the FM in subjects with severe CMI. The result of a shorter fastigium height in CMI patients has been reported in the literature [14, 18, 19]. The Boogard angle showed the opposite trend to those for PCF area, clivus length, and fastigium height. The Boogard angle tends to increase with an increase in tonsil length, with the highest horizontal angulation found in group CMI > 5. Only two comparisons showed significant differences in the Boogard angle (CON−TP vs. CMI > 5 and CON+TP vs. CMI > 5). No significant differences were found in the comparison of the groups with low-lying TP (CMI < 5 vs. CON+TP), the two control groups (CON−TP vs. CON+TP), and the two CMI groups (CMI < 5 vs. CMI > 5). This result agrees with those of Dufton et al. and Houston et al., where a larger Boogard angle was reported for CMI subjects than for controls [11, 14]. With a larger Boogard angle, one would expect that the PCF area will be larger since a more horizontally angled clivus might provide more space posterior to the clivus. However, a correlation between the Boogard angle and the PCF area suggests otherwise (see Fig. 5a). A weak negative correlation was found between the Boogard angle and PCF area (r = − 0.12, p = 0.04), suggesting that the PCF area decreases with an increase in the Boogard angle. Other morphometric parameters, including the clivus length (r = − 0.46, p = 4e⁻¹⁷) and posterior CSF area (r = − 0.30, p = 1e⁻⁶), also showed a negative correlation with the Boogard angle (see Fig. 5 b and d).

Fig. 5.

Correlation between the Boogard angle and a PCF area, b clivus length, c anterior CSF area, and d posterior CSF area. The correlation was performed by combining all the CMI and control groups

We also evaluated the areas of the anterior and posterior CSF space, as these two parameters may provide a measure of soft tissue structure crowding in the region inferior to the FM. The anterior and posterior CSF areas followed a pattern similar to some of the morphometric parameters described above, as an increase in the tonsil length was associated with a decrease in anterior and posterior CSF areas (see Fig. 3d). The anterior CSF area was found to be similar between groups with low-lying tonsils (CON+TP vs. CMI < 5), the CMI groups (CMI > 5 vs. CMI < 5), and the control groups (CON−TP vs. CON+TP). This finding implies that although the anterior CSF area decreases with a larger tonsillar descent, having a larger tonsillar descent will not necessarily lead to a smaller anterior CSF area. For the posterior CSF area, an increase in the tonsil length may decrease the area occupied by CSF, resulting in a smaller CSF area and possibly an increase in crowding of the posterior region caudal to the FM. This assumption is strengthened by the fact that we found a significant difference between CMI individuals having severe tonsillar descent (CMI > 5) as compared to symptomatic CMI individuals with low-lying tonsils (CMI < 5).

Clinical implications

In this study, we compared several SRCPs for individuals with CMI < 5 and CMI > 5 to determine any similarities/differences between the two CMI groups. While we found several morphometric differences between the two groups (including clivus length, fastigium height, and posterior CSF area), the SRCPs of the two groups were similar. Additionally, there were no significant correlations between TP and SRCPs. This result supports the growing consensus that TP is not indicative of CMI symptom severity [24]. In contrast to our findings, Kavak et al. found that the amount of tonsillar herniation was positively correlated with the modified Asgari score [21].

Although TP did not correlate significantly with any SRCPs, other morphometric parameters did. For example, we found that CMI individuals with TP < 5 and larger anterior CSF area may be less likely to suffer from anxiety (r = − 0.37, p = 0.03). In addition, for individuals with severe CMI (TP > 5), the shorter the clivus length, the more likely it is for a patient to report weakness (r = − 0.33, p = 0.009) and hypermobility (r = − 0.36, p = 0.004).

Clivus length was the only morphometric parameter that showed a significant correlation with weakness and hypermobility in the CMI > 5 group. We also found a trend between clivus length and double vision in the CMI > 5 group. Interestingly, the abducens nerve (cranial nerve VI), the longest intracranial nerve and the one responsible for ipsilateral eye abduction [34], passes through the Dorello canal, which is posterior to the clivus. The long intracranial route of this nerve makes it susceptible to damage due to aneurysms, fractures, or increased intracranial pressure [35], which may lead to bilateral abducens nerve palsy (often present with double vision, [34]), a condition reported in CMI patients [35–37]. Changes in the shape of the clivus as reported in CMI individuals may result in changes in the abducens nerve, which in turn may increase the chances that CMI individuals will experience double vision. In the current study, symptom severity did not correlate with any morphometric parameters in either the CMI < 5 group or the CMI > 5 group. This finding agrees with the results of an analysis by Houston et al., in which 24 out of 27 morphometric parameters did not correlate with symptom severity [9]. In addition, Bolognese et al. reported the opinions from a panel of 63 international experts in CMI. Over 85% of respondents agreed that the 5 mm rule for CMI diagnosis should be rejected. Moreover, 95.1% of respondents agreed that CMI is characterized by tonsillar herniation; however, a tonsillar herniation may result from disease processes or mechanisms other than tonsillar herniation [24]. Our results further support the assertion that the 5 mm cutoff is arbitrary and should not be used in, and of itself as an indication for CMI symptomatology.

This study was limited by a retrospective design and the clinical findings were based solely on self-reported data provided through the Chiari1000 database by CMI patients. While the self-reported scales have been validated extensively, they have not been validated specifically for CMI subjects with significant or minor tonsillar descent. These scales were included to be sure we were not missing any large changes that might be detected. However, future studies that apply these scales prospectively at the time of diagnosis may find different results. Past CMI research that we and others have published have shown that these questionnaires do correlate with pain (e.g., Allen et al., 2014, 2018; Garcia., 2018, 2019; Houston et al., 2018, 2019, [9, 14, 20, 38–40]) Additionally, the average time between the image acquisition and participation in the Chiari1000 survey was 2.3 ± 2.7 years. Thus, there was a possibility of inaccurate reporting of the clinical symptoms due to poor recall. However, we do not expect recall bias by the participants in the Chiari1000 with respect to the morphometric parameters. The use of midsagittal images for determination of the TP may also be a source of error, as different results may be obtained if the measurements were obtained off-midsagittal plane. Furthermore, the measurements were performed on the midsagittal plane of T1 and T2 images, the differences in sequences, or performing the measurements off midsagittal may have an impact on our morphometric analysis. However, only seven T2-images (2% of the total number of images) were present in this study, and thus, we did not expect these to have any impact on our results or conclusions. Also, some of the morphometric parameters demonstrated moderate collinearity, and thus, may have an impact on our final results. The lack of a sufficient number of male and juvenile image sets in the Chiari1000 database prevented us from including these groups in our study.

Conclusions

We found significant morphometric differences between adult female individuals having different levels of TP. Even though morphometric differences were found between CMI groups with TP above and below 5 mm, a comparison of the SRCPs of these groups revealed no significant differences. In addition, we showed that for adult female individuals with symptomatic CMI as compared to incidentals with similar TP, the only morphometric parameter that can reliably distinguish the two groups was the clivus length. Thus, the clivus may be able to distinguish between adult female incidentals with low-lying tonsils and adult female symptomatic individuals with CMI. However, given the constraints of this study (retrospective, adult patients, and female gender) a more generalized inference cannot be derived at this time. We hope to expand the scope of this study to include pediatric and male participants. Ultimately, it should be emphasized that the decision to operate on patients with 1–5 mm descent lies in the clinical evaluation of the patients by the neurosurgeon, and not on radiological findings alone.

Future work may focus on extending the finding of the clivus to incidentals with TP > 5 mm and exploring how these incidentals differ from CMI individuals with TP > 5 mm. Additionally, another area of interest is the use of tonsillar shape to discriminate between symptomatic CMI and incidentals. Currently, in the literature, there are no good mathematical descriptions to evaluate the importance of tonsillar shape. We will continue to examine this in future studies. Finally, we have shown that while TP is not indicative of CMI symptomatology, other brain morphometrics such as clivus length may have some predictive utility. The significance of the shortened clivus to CMI anatomy and CMI symptoms should also be a focus of future investigation.