LSTV was first regarded as a cause of low back pain by Bertolotti in 1917 [12]. Although there have been many studies on LSTV, a lot of misunderstandings about its anatomical characteristics existed. Some anatomical landmarks suggested previously for determining spinal counts and LSTV were verified unreliable [5, 13], while were still taken as criteria by some other studies [14, 15].
Castellvi Classification System was widely accepted for its simplicity. Prominent characteristics of each Castellvi type of MA-LSTV were initially extracted from Ferguson view of LPR [3]. Considering that radiographic image forms as a result of differing attenuation of the x-ray beam by various tissues within the patient, characteristics of concerned bony structures might be disturbed or covered by other surrounding structures. Recently, Farshad-Amacker et al reported that coronal MRI is superior to standard AP-LPR in detecting and classifying of MA-LSTV [10]. However, MRI is inferior in detecting bony details compared with CT-CRIs. Moreover, coronal MRI is not routinely taken in routine clinical practice.
In our series, we found all suspected MA-LSTVs diagnosed by AP-LPR were verified to be true MA-LSTVs by CT-CRIs, however, AP-LPR could not classify MA-LSTV types with 100% accuracy. Type IIIb MA-LSTV might be wrongly classified as type IV or IIb, while type IV and IIIa wrongly classified as type IIb and IIa respectively. The agreement of classification between LPR and CT groups was poor to moderate according to the statistical analysis.
What caused such misclassification? The following were our analysis.
1. BUS might not occupy the full space between the anomalous TP and sacrum at the type III transitional side.
According to Castellvi classification principle, if continuous bone bridge forms between the TP of MA-LSTV and sacrum, the anomalous side belongs to a type III transition. One might take it for granted that bone bridge formation means that all space between the TP and sacrum should be fully occupied by continuous bone with disappearance of RSB. In reality, such condition really exists (Fig. 3, left side). But this only occurred in 48.8% (39/80) of type III sides in our series, other situations might also exist. For example, only limited percent of the space might reach bony fusion while the remnant was occupied by JLS, under such condition, the abnormal side was a type III transition, but there was a high possibility that the side was misinterpreted as a type II side by AP-LPR. This could be further divided into 3 subcategories: (1) BUS only occupied limited region while JLS occupied most region (Fig. 1, left side); (2) BUS occupied most region while JLS occupied limited region (Fig. 2, right side); (3) JLS and BUS shared almost equal space (Fig. 2, left side). These situations occurred in 31.3% (25/80) type III transitional sides, and occupied 61.0% (25/41) of all misclassified sides.
2. Existence of RSB might lead a type III transitional side misdiagnosed as type II transition side on AP-LPR.
In some cases, complete fusion between TP from MA-LSTV and sacrum had reached, while irregular RSB at the inferior boundary of TP from MA-LSTV and/or that at the superior boundary of sacrum still existed. This might form a false image of anomalous articulation on AP-LPR, resulting in a type IIIa transition being misdiagnosed as a type IIa transition (Fig. 3–4). If discrepancy of width at the connection space existed at bilateral sides in a type IIIb case, there was a higher possibility that the narrower side with RSB was misdiagnosed as a type II transition. These situations occurred in 20.0% (16/80) of type III sides and occupied 39.0% (16/41) misdiagnosed sides.
3. There might exist a progressive transformation process from type II transition to type III transition.
In our series, we found type III transitional side in many cases was consisted of both JLS and BUS at connection space instead of only occupied by BUS, the former occurred in 31.3%(25/80) of all type III sides, verified by CT-CRIs, while the distribution area or percentage of JLS or BUS varied in different cases. In the remaining 55 sides, 16 were found existing RSB, this further occupied 19.8% (16/80) of type III transitional sides. Previously, we had taken it for granted that there only existed two clear categories at the space between abnormal TP of MA-LSTV and sacrum: one situation was that all space was replaced by continuous bone connection with complete disappearance of RSB, another was that all space was replaced by abnormal articulation structure. No transitional process existed between these two definite categories. When the reality revealed there existed the above-mentioned situations, that is, both JLS and BUS co-existed at the connection space to different distribution, or RSB left at the connection region, we hypothesized boldly that type IIa transitional side might develop into type IIIa transitional side under some special conditions. Initially, one type II transitional side might develop to the stage that bony connection developed only at limited region. Gradually, more JLS was replaced by bony bridge, till all connection space was replaced by bony bridge but with RSB remaining. At final stage, RSB disappeared with complete rigid bony connection. Recently, Hou et al reported one case who developed type IIIa MA-LSTV from type IIa following discectomy and fusion at lumbosacral level [16]. This phenomenon partially supported our hypothesis. However, to verify this assumption, further follow-up on MA-LSTV with CT examination was needed. Also, if the hypothesis was true, the type of MA-LSTV should not be thought as congenital, but acquired, at least in some cases.
4. AP view instead of Ferguson view LPR was used for MA-LSTV type classification.
In Castellvi’s original literature, 30°angled AP view (Ferguson view) of lumbar spine was regarded as true AP of lumbosacral joint, which could reach the purpose of removing the radiographic overlap of abnormal TP from MA-LSTV on sacrum [3,4,10]. This viewpoint might be based on the following assumptions: (1) 30° angled AP view of lumbar spine indeed was the true AP of lumbosacral joint; (2) coronal plane of abnormal TP form MA-LSTV was perpendicular to inferior endplate of MA-LSTV; (3) inferior edge of abnormal TP from MA-LSTV was parallel to inferior endplate of MA-LSTV; (4) the cleft between TP of MA-LSTV was parallel to coronal plane observed from transaxial view. However, Ferguson view is not routinely taken as the standard radiographic assessment of lumbar spine in clinical practice. As a retrospective study, we took AP view instead of Ferguson view LPR for MA-LSTV type classification. This might result in more misclassifications. However, according our image data, we found the orientation of JLS is irregular, neither parallel to horizontal plane on AP view, nor to coronal plane on transaxial view, nor to endplate on oblique view (Fig. 5), which was theoretically impossible to offset by Ferguson view LPR. A recent study indicated that Ferguson view had no superiority over the standard AP pelvis view for grading of sacroiliitis [17]. Further study might be needed to compare the AP view with the Ferguson view of the LPR in order to resolve whether one modality has a clear advantage for classification of MA-LSTV types.
Previously, conclusions of relationship between various MA-LSTV types and their clinical significance were based on suspected MA-LSTV classification identified by LPRs, which might be questionable. The real relationship should be re-evaluated based on real MA-LSTV types identified by CT-CRIs.