Retro-Odontoid Pseudotumor in Atlantoaxial Instability: Insights Into Presence, Subtypes, and Postoperative Regression
Article information
Abstract
Objective
Retro-odontoid pseudotumor (ROP) is a nonneoplastic mass associated with atlantoaxial instability (AAI). This study compared ROP-positive and ROP-negative AAI patients and evaluated cystic versus granulation-type ROP regarding regression patterns and surgical outcomes.
Methods
We retrospectively analyzed 112 AAI patients who underwent pre- and postoperative imaging and clinical evaluations. Patients were classified as ROP-positive or ROP-negative, with ROP-positive cases further categorized as cystic or granulation-type. Imaging parameters—including atlantodental interval (ADI), ΔADI, and cervical range of motion (ROM) were compared along with regression time and postoperative outcomes.
Results
Among 112 patients, 57 (50.9%) had ROP. The ROP-positive group was older (67.37±13.13 years vs. 56.90±15.15 years, p<0.001) and had lower ADI (5.63±2.77 mm vs. 6.99±2.33 mm, p=0.034), ΔADI (3.01±2.27 mm vs. 3.89±2.07 mm, p=0.006), and C2–7 ROM (30.78°±15.45° vs. 41.73°±16.58°, p<0.001). In ROP subgroups, the cystic group had greater C1–2 ROM (15.69°±6.34° vs. 10.00°±7.72°, p=0.013) and ADI (6.98±2.68 mm vs. 5.14±2.66 mm, p=0.042). Immediate postoperative ROP thickness remained greater in the cystic group (6.85±2.49 vs. 5.21±1.82 mm, p=0.042), while ROP thickness at 3 months and 1 year showed no significant differences. JOA recovery rates were similar.
Conclusion
This study demonstrates that ROP-positive AAI patients exhibit distinct radiological characteristics, with reduced cervical mobility. Furthermore, cystic ROP shows delayed regression following posterior fusion. These findings underscore the importance of ROP subtypes in surgical planning, requiring closer monitoring and possibly earlier intervention.
INTRODUCTION
Retro-odontoid pseudotumor (ROP), also known as periodontoid pseudotumor or pannus, is a nonneoplastic soft tissue masses adjacent to the odontoid process of C2, which can cause cervicomedullary compression [1]. Acute inflammation in ROP can manifest neck pain or headache. As a chronic process, mass effect on the cervical spine can manifest as myelopathy including sensory and motor deficits [2,3].
Various factors have been identified as contributors to the development of ROP. It has been established that ROP is associated with repetitive and chronic ligament damage related to the atlantoaxial joint [3].
Currently, screw-based constructs provide rigid stabilization and improve surgical outcomes in various pathologies around the craniovertebral junction [4-9]. However, there has been controversy regarding the surgical treatment of ROP. The treatment methods can be divided into direct removal of ROP or posterior fusion with or without decompression [10]. In the removal of ROP, the posterior transdural approach has the drawback of a higher risk of neurological damage, therefore, the safer transoral approach has been prioritized [11,12]. Nevertheless, this method still has shortcomings, in the ocurrence of cerebrospinal fluid leakage, wound infection, arterial injury, and pharyngeal wound dehiscence [13]. On the other hand, it has been reported by Grob et al. [14] that posterior decompression and fusion not only provide neurological decompression but also reduce ROP. As similar results continue to emerge, treatment for myelopathy associated with ROP has established upper cervical fusion as more common treatment [3,15-18]. While posterior fusion is the preferred surgical approach, the variability in regression patterns of ROP raises questions about optimal treatment strategies.
Previous studies have been reported on radiologic features which affect ROP regression following upper cervical fusion, such as cystic degeneration and calcification. However, no consensus exists on the relationship between these characteristics and clinical outcomes.
Therefore, this study aims to (1) compare the characteristics of AAI patients with and without ROP, (2) investigate the radiological and clinical features of cystic versus granulation-type ROP, and (3) analyze the regression patterns of these subtypes following posterior fusion.
MATERIALS AND METHODS
This study is a retrospective study conducted on patients diagnosed and treated for atlantoaxial instability (AAI) at 2 independent institutions from 2007 to 2023. After obtaining approval and a waiver of informed consent from the Institutional Review Board (IRB) of the Catholic University of Korea (IRB No. XC22 DDDT0060), patient data were analyzed.
The inclusion criteria were as follows: patients aged 18 years or older at the time of admission and diagnosed with AAI, with or without ROP. Only patients who had preoperative and postoperative x-rays, including flexion and extension views, as well as magnetic resonance imaging (MRI) evaluations conducted before surgery and immediately after surgery, were included. For surgically treated patients, only those with available preoperative and postoperative Japanese Orthopaedic Association (JOA) scores and JOA recovery rates (RR) were included in the study.
AAI was defined as an atlantodental interval (ADI) greater than 3 mm, measured as the distance between the posterior aspect of the anterior arch of the atlas and the anterior aspect of the odontoid process [19]. In this study, ROP was defined as a lesion in the retro-odontoid space, appearing hypointense on T1-weighted images and hypointense to hyperintense on T2-weighted images on sagittal MRI, with a thickness of 4 mm or more. This threshold is based on previous studies indicating the average anteroposterior thickness of the transverse atlantal ligament is 3.4 mm [20]. In addition, as described by Nakano et al. [21], cystic degeneration was defined as a lesion located posterior to and clearly distinguishable from the ROP, appearing isointense to hypointense on T1-weighted and hyperintense on T2-weighted sagittal images.
Preoperative x-rays were used to assess the ADI and the change in ADI (ΔADI) between flexion and extension. Additionally, Cobb angles and range of motion (ROM) between C0 and C1, C1 and C2, and C2 and C7 were evaluated. To examine the impact of basilar invagination (BI), the Redlund-Johnell index and Ranawat index were measured [22,23]. Preoperative computed tomography scans were utilized to determine the presence of dens erosion and ROP calcification, while the degree of arthritis in the atlantoaxial facet joint was assessed using the grading system reported by Shimizu et al. [24] Furthermore, MRI scans were conducted preoperatively, immediately postoperatively, and at 3 months and 1 year postoperatively to evaluate ROP regression and to measure the space available for the cord (Fig. 1). The immediate postoperative MRI was performed on the day the Jackson-Pratt drain, which had been inserted intraoperatively, was removed. The drain removal occurred between 2 and 5 days postoperatively (mean±standard deviation: 2.96±0.76 days). MRI scans were conducted using T1-weighted, T2-weighted, and short tau inversion recovery sequences.
Measurement of space available for the spinal cord (SAC). SAC was measured using a sagittal image from magnetic resonance imaging T2-weighted image. The distance was measured between the posterior border of retro-odontoid pseudotumor (ROP) and the anterior border of C1 posterior arch at the midline preoperatively while it was measured from the ROP to the margin where the posterior cerebrospinal fluid signal ends postoperatively.
Exclusion criteria included cases where follow-up data were not available for up to 1 year postoperatively, patients who had undergone prior high cervical surgery, and those with missing imaging data necessary for analysis.
Where surgery was performed, occipitocervical fusion or atlantoaxial fusion with or without C1 laminectomy were employed according to the pathologic lesion, and all procedures were conducted by the same senior surgeon.
The collected data were statistically analyzed using IBM SPSS Statistics ver. 20.0 (IBM Co., USA). For categorical data, the chisquare test was used, while for continuous variables, independent t-tests and paired t-tests were employed to assess differences between the 2 groups. A p-value of 0.05 or less was considered statistically significant.
RESULTS
1. Comparison Between ROP-Positive and ROP-Negative Groups in AAI
A total of 112 patients with AAI were identified, and ROP was observed in 57 of these patients. When comparing the ROPpositive and ROP-negative groups, the ROP-positive group was found to be older (ROP-negative: 56.90±15.15, ROP-positive: 67.37±13.13; p<0.001). However, no statistically significant differences were observed between the 2 groups in terms of gender distribution or associated etiologies (Table 1).
Demographic details in all patients with AAI
The differences in radiologic features between the 2 groups were examined. In the ROP-negative group, ADI (ROP-negative: 6.99±2.33 mm, ROP-positive: 5.63±2.77 mm; p=0.006), ΔADI (ROP-negative: 3.89±2.07 mm, ROP-positive: 3.01±2.27 mm; p=0.034), C2–7 ROM (ROP-negative: 41.73°±16.58°, ROP-positive: 30.78°±15.45°; p<0.001) were significantly larger. On the other hand, the C1–2 angle was greater in the ROP-positive group (ROP-negative: 18.04°±13.53°, ROP-positive: 25.84°±9.43°; p=0.002) (Table 2).
Radiologic parameters between ROP-negative and ROP-positive groups
2. Comparison Between ROP Patients With and Without Cystic Degeneration
In 57 patients with ROP, cystic degeneration was identified in 14 patients, while only granulation tissue in ROP was radiologically observed in 43 patients. Based on these findings, the patients were classified into 2 groups—the cyst group and the granulation group—and the radiological characteristics and regression patterns following posterior cervical fusion were analyzed. One case in the cyst group underwent direct surgical removal of the cyst, which did not represent spontaneous regression following fusion, and was therefore excluded from the analysis.
The analysis revealed a significant difference in sex distribution between the 2 groups: 2 out of 13 patients were male (15.38%) in the cyst group, whereas 22 out of 43 patients were male (51.16%) in the granulation group (p=0.022). In addition, os odontoideum was more frequently identified in 7 patients (53.85%) in the cyst group, compared to 4 patients (9.30%) in the granulation group. On the other hand, enthesopathy, encompassing ossification of the posterior longitudinal ligament, ossification of the anterior longitudinal ligament, ankylosing spondylitis, and diffuse idiopathic skeletal hyperostosis, as well as rheumatoid arthritis (RA), BI, and atlanto-occipital assimilation (AOA), were more frequently observed in the granulation group. The respective frequencies in the cyst group versus the granulation group were 7.7% versus 16.28%, 7.7% versus 16.28%, 0% versus 11.63%, and 0% versus 13.95%, in the order mentioned. The overall distribution of etiologies showed a statistically significant difference between the 2 groups (p=0.013). The proportion of cases in which interfacetal spacers were inserted into the C1–2 facet joint following C1 laminectomy for fusion enhancement was compared between the groups. In the cyst group, 6 of 13 patients (46.2%) underwent the procedure, whereas in the granulation group, 30 of 43 patients (69.8%) did. However, this difference was not statistically significant (p=0.220) (Table 3).
Details of ROP-positive patients with AAI
The granulation group showed significantly lower ADI (5.14±2.66 mm vs. 6.99±2.33 mm, p=0.006), ΔADI (2.68±2.03 mm vs. 3.89±2.07 mm, p=0.034), and C2–7 ROM (29.04°±14.80° vs. 41.73°±16.58°, p<0.001) compared to the ROP-negative group, while the C1–2 angle was significantly greater (25.84°±9.43° vs. 18.04°±13.53°, p=0.002). However, when contrasted with the ROP-negative group, the cyst group did not show statistically significant differences in any radiologic parameters, including ADI, C1–2 angle, and C2–7 ROM (Table 4).
Comparison of radiologic parameters among ROP-negative, ROP-positive cyst, and ROP-positive granulation groups
Compared to the granulation group, the cyst group exhibited significantly greater C1–2 ROM (15.69°±6.34° vs. 10.00°±7.72°, p=0.013) and ADI (6.98±2.68 mm vs. 5.14±2.66 mm, p=0.042), while no significant differences were observed in C2–7 ROM, C2–7 angle, or other radiologic parameters between the 2 groups (Table 4). On the other hand, the reduction in ROP thickness immediately after surgery was less prominent in the cyst group, as supported by a greater remaining thickness compared to the granulation group (cyst group: 6.85±2.49 mm, granulation group: 5.21±1.82 mm, p=0.042) (Table 5). However, ROP thickness showed no significant difference between the 2 groups as time progressed, at 3 months postoperatively (cyst group: 3.46±2.04 mm, granulation group: 3.83±1.69mm, p=0.568) and 1 year postoperatively (cyst group: 2.83±1.71 mm, granulation group: 3.10±1.79 mm, p=0.574) (Figs. 2 and 3). Accordingly, when JOA RR was measured before surgery and 1 year after surgery, both groups showed improvement in myelopathy symptoms (cyst group: 57.38%±22.74%, granulation group: 59.36%±27.36%, p=0.711) (Table 5).
Differences of postoperative outcomes between cyst and granulation group
Regression pattern of retro-odontoid pseudotumor (ROP) in cyst group. POD, postoperative day; SD, standard deviation; Preop, preoperative; Postop, postoperative.
Regression pattern of retro-odontoid pseudotumor (ROP) in granulation group. POD, postoperative day; SD, standard deviation; Preop, preoperative; Postop, postoperative.
3. Illustrative Cases
1) Case 1 (early regression of ROP)
An 85-year-old male patient presented with progressive headache and worsening quadriparesis over the past 6 months. MRI revealed granulation-type ROP causing cord compression, accompanied by AAI. Posterior decompression of the cord was performed via C1 laminectomy, and to address the instability, lateral mass screws were placed in C1 and pedicle screws in C2 for fixation. Additionally, 5-mm cervical spacers were bilaterally inserted into the atlantoaxial joint to promote fusion. Postoperatively, the patient experienced improvement in hand numbness and a dramatic reduction in headaches. An MRI taken 5 days after surgery showed early regression of ROP thickness (Fig. 4A and B).
Representative cases of ROP regression. (A and B) Early regression of granulation-type retro-odontoid pseudotumor (ROP). (A) Preoperative sagittal T2-weighted magnetic resonance imaging (MRI) shows a granulation-type ROP causing significant cord compression (white arrow). (B) Postoperative MRI taken 5 days after surgery demonstrates marked regression of ROP thickness (red arrow) following C1 laminectomy, C1-2 fusion, and intraarticular spacer insertion. (C–E) Delayed regression of cystic-type ROP. (C) Preoperative sagittal T2-weighted MRI reveals a cystic-type ROP with cord compression (white arrow). (D) Postoperative MRI at 10 days shows no significant change in ROP thickness and cord compression (red arrow). (E) However, MRI at the 1-year follow-up demonstrates significant regression of the cystic ROP, correlating with clinical improvement.
2) Case 2 (delayed regression of ROP)
A 76-year-old female patient underwent MRI after experiencing abnormal gait and weakness in both hands for 1 year. She was diagnosed with AAI accompanied by the cystic-type ROP. C1 laminectomy and C1–2 fusion were performed with the insertion of cervical spacers into the atlantoaxial joints. Postoperatively, the sensation in her legs improved compared to preop-eratively, but no improvement was noted in hand weakness or sensory abnormalities immediately after surgery. However, the patient’s myelopathy symptoms improved from 9 to 15 in JOA score 1 year after surgery. Radiologically, postoperative MRI showed no regression in the thickness of the cystic-type ROP, but a follow-up MRI taken at one year showed significant regression of the ROP (Fig. 4C–E).
DISCUSSION
Research on ROP regression following posterior fusion has been continuously published [1,10,15,16,24-28]. However, there is only one research on the factors that lead to rapid ROP regression identifying that older age and thicker preoperative ROP were associated with faster regression after posterior fusion using retrospective data of 11 patients [25]. Therefore, further research is needed to explore other factors involved in the regression pattern of ROP based on preoperative imaging characteristics.
In the present study, the authors compared the characteristics between the ROP-positive and the ROP-negative AAI groups. The ROP-positive group showed a tendency to be older, which may be related to research findings indicating that the occur-rence of ROP is associated with increasing age [29-31]. Additionally, the demographic analysis revealed no significant differences in the distribution of etiology. Although there have been reports about ROP related conditions such as RA, congenital anomalies, os odontoideum, or trauma, the analysis of the etiological distribution between ROP-positive and ROP-negative groups in AAI patients has not been previously conducted, which makes it an intriguing finding in this study (Table 1) [15,26,32].
In the ROP-negative group, it was observed that not only ΔADI but also C2–7 ROM were greater compared to the ROPpositive group, suggesting that the ROP-negative group exhibited a relatively more mobile motion (Table 2). According to Niwa et al. [15], an analysis of non-rheumatoid ROP patients indicated that restriction in subaxial ROM contributes to the development of ROP. Furthermore, their study showed a negative correlation between ROP thickness and both ADI and subaxial ROM, supporting the finding that ROP is more frequently observed in stiffer spines in our study.
Meanwhile, Shin et al. and Goel et al. mentioned cystic degeneration associated with ROP and reported that posterior decompression and fusion alone can lead to spontaneous regression of both the ROP and the cyst within 6 to 12 months [21,26,27]. However, the patients with ROP were divided into 2 groups in this study—the cyst group and the granulation group—to investigate the clinical significance of cystic degeneration and its clinical and radiologic features.
In this analysis, a clear difference in the distribution of etiologies was observed between the 2 groups. The cyst group showed a higher prevalence of os odontoideum, whereas enthesopathy, RA, BI, and AOA were more commonly seen in the granulation group (Table 3). Additionally, the cyst group demonstrated greater C1–2 ROM and ADI compared to the granulation group, while ROP calcification was more prominent in the granulation group (Table 4). These findings support the hypothesis that cystic-type ROP arises from severe biomechanical stress at the C1–2 junction, where chronic instability and repetitive subclinical hemorrhage within the transverse ligament trigger a pathogenetic cascade resembling synovial cyst formation. This parallels degenerative adaptations observed in other spinal regions (e.g., lumbar synovial cysts), reinforcing the concept of shared instability-driven pathways in cystic ROP pathogenesis [33,34].
Based on the findings that the ROP-negative group had greater ΔADI and C2–7 ROM than the ROP-positive group, and that the cyst group exhibited higher C1–2 ROM and ADI than the granulation group, it can be inferred that both the ROP-negative and cyst groups represent mobile phenotypes. This observation led us to perform an additional analysis to determine whether radiologic differences similar to those observed between the ROP-negative and ROP-positive groups also exist between the ROP-negative and cyst groups.
Interestingly, no significant radiologic differences were found between the ROP-negative and cyst groups. Our findings propose 2 mechanistic frameworks to explain the pathomechanistic duality of ROPs. First, cystic and granulation-type ROPs, while both associated with C1–2 instability, likely represent distinct pathological entities driven by opposing biomechanical processes (Fig. 4). Cystic ROPs arise from chronic instabilitydriven synovial fluid leakage, repetitive microbleeding, or degenerative cystic changes in connective tissues. In contrast, granulation-type ROPs reflect a compensatory stabilization response, characterized by transverse ligament hypertrophy and fibrocartilaginous proliferation to counteract segmental instability. This dichotomy is reinforced by the divergent distributions of underlying cervical pathologies: cystic ROPs correlate with subaxial spinal stiffness, which amplifies C1–2 stress, while granulation-type ROPs associate with focal degenerative osteoarthritis (Table 3).
Second, we hypothesize a potential temporal progression model, where cystic ROPs may represent an intermediate stage between ROP-negative AAI and solid granulation-type pseudotumors. Repetitive stress at the C1–2 junction could initially induce cystic degeneration, with subsequent fibrotic remodeling culminating in granulation mass formation. However, longitudinal evidence for this sequence remains elusive in current literature, necessitating prospective studies with serial imaging to validate causality.
Another interesting finding is the postoperative ROP regression pattern. In the granulation group, ROP thickness decreases immediately postoperatively and continues to show a gradual reduction over time. In contrast, the cyst group shows no significant change in the immediate phase, with a notable decrease only starting around 3 months after surgery (Figs. 2 and 3). Delayed postoperative regression of cystic ROP compared to granulation-type lesions carries critical implications for surgical strategy. While both subtypes may regress following posterior fixation, cystic ROP’s slower resorption timeline—often showing no significant reduction until 3-month postsurgery—suggests that indirect stabilization alone may inadequately address acute neurological deterioration. While posterior fixation addresses the instability underlying both ROP subtypes, cystic ROP’s delayed regression necessitates a tailored approach. Surgeons should consider direct decompression in cystic ROP cases with significant cord compression to mitigate neurological risks during the prolonged resorption period. In this context, C1 laminectomy or direct cyst removal may lead to better outcomes (Fig. 5).
A representative case of direct retro-odontoid pseudotumor (ROP) cyst resection with favorable outcome. (A–D) Direct excision of the cystic ROP. Preoperative sagittal (A) and axial (C) T2-weighted MRI images reveal a cystic ROP associated with severe cord compression (white arrow). Postoperative sagittal (B) and axial (D) MRI images show complete resolution of the cystic ROP following epidural cyst removal with significant clinical recovery (red arrow).
Another noteworthy aspect of ROP regression patterns is the inconsistency between our findings and those of previous studies. For instance, Nakano et al. [21] reported greater regression in ROPs with accompanying cysts at 6 months postoperatively, whereas our study found no significant difference in ROP thickness between the cystic and noncystic groups at 3 months after surgery.
This discrepancy may be attributable to methodological differences between the studies. Specifically, Nakano et al. [21] included a smaller sample size, and the preoperative ROP thickness was notably greater in the cystic group compared to the noncystic group, potentially introducing bias and limiting the generalizability of their findings. Furthermore, differences in the timing of postoperative assessments (3 months vs. 6 months) may also have influenced the observed regression patterns.
Further research with larger cohorts and longer follow-up periods is warranted to clarify the temporal dynamics of ROP regression, particularly in relation to cystic changes. Prospective, longitudinal studies that control for baseline lesion size and utilize standardized imaging intervals will be essential to better understand the natural history and optimal management strategies for different ROP subtypes.
This study has several limitations that warrant consideration when interpreting the findings. First, the retrospective design introduces the potential for selection bias, as patient inclusion was dependent on the availability of complete clinical and radiological data. Second, the sample size, particularly within the cystic ROP subgroup (n=13), is relatively small. This limited sample size may have reduced the statistical power to detect significant differences between the cyst and granulation groups, particularly for secondary outcomes. Future studies with larger cohorts are needed to confirm these findings and to assess the clinical relevance of ROP subtype on long-term functional outcomes.
Third, the study period spans a considerable length of time (2007–2023), during which surgical techniques and postoperative management strategies may have evolved. While all surgeries were performed by the same senior surgeon, subtle variations in approach or instrumentation over time could have influenced ROP regression patterns.
Fourth, the study lacks long-term follow-up beyond one year postoperatively. While significant ROP regression was observed in both groups, the long-term stability of these changes and their impact on clinical outcomes remain unclear. Furthermore, the study does not account for potential confounding factors such as smoking status, body mass index, or other comorbidities, which may influence ROP regression and overall health status.
Fifth, heterogeneity in surgical methods may potentially hinder the ability to draw meaningful conclusions. In some patients, interfacetal cages were inserted into the bilateral C1–2 facet joints following C1 laminectomy to compensate for the reduced fusion bed. Although the difference in the use of interfacetal cages between the cyst group and the granulation group was not statistically significant (p=0.220), previous reports have shown that ROP thickness can decrease following C1 laminectomy alone [35,36], and the insertion of an interfacetal cage may lead to improving alignment and stretching of the ROP. Therefore, future studies with standardized surgical techniques are warranted.
Finally, while the JOA score is a widely used outcome mea-sure for cervical myelopathy, it may not fully capture the nuances of functional recovery and quality of life. Future studies should consider incorporating additional patient-reported outcome measures, such as the Neck Disability Index or the 36-item Short Form health survey, to provide a more comprehensive assessment of treatment effectiveness.
Despite these limitations, this study provides valuable insights into presence, subtypes, and postoperative regression patterns of ROP in AAI patients. This study highlights the importance of understanding the distinct regression patterns of ROP subtypes, suggesting that surgical strategies should be tailored based on radiologic features and severity of ventral compression. Future studies with larger sample sizes and longer follow-up periods will be essential to refine treatment guidelines and determine the optimal management approach for this complex condition.
CONCLUSION
This study underscores the distinct radiological differences between ROP-positive and ROP-negative groups, as well as between cystic and granulation-type ROP groups. The findings suggest that restricted cervical mobility, characterized by reduced ADI and ROM, is associated with the development of ROP, and that cystic degeneration tends to occur in cases of greater instability, particularly in the presence of os odontoideum. The regression patterns of ROP differ based on the type, with the cystic group showing a delayed reduction in ROP thickness compared to the granulation group. These radiologic insights can guide surgical decision-making, particularly regarding the method of decompression and fusion and suggest that surgical strategies should be tailored based on ROP subtypes. Future studies should focus on analyzing regression patterns across different surgical approaches to optimize treatment outcomes.
Notes
Conflict of Interest
The authors have nothing to disclose.
Funding/Support
This study received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
Author Contribution
Conceptualization: DHK, JTH, ISK, JBL, HJL; Formal analysis: DHK, JWH, ISK, JYK, JJL, JBL; Investigation: DHK, JWH, JYK, JJL, HJL; Methodology: DHK, JTH, JWH, ISK, JJL, JBL; Project administration: DHK, JTH, JYK, HJL; Writing – original draft: DHK; Writing – review & editing: JTH, ISK, HJL.