Statistical analysis

Analyses were done in IBM SPSS Statistics 23.0 (IBM, Armonk, NY, USA). Patient characteristics were analyzed using frequencies. Investigating proportions between unpaired groups of categorical data was done with Chi Square test. Comparison of measurements of the spinal canal between CES patients and lumbar herniated disc patients with sciatica and without CES was done with Mann-Whitney U test.

To evaluate the effect of MRI features on clinical presentation and outcome, binary logistic regression models were built, with MRI features as independent variables (severity of cauda equina compression; type of cauda equina compression i.e. unilateral or bilateral; level of disc lesion) and clinical features as dependent variable. Since there were 4 clinical features (presence of micturition dysfunction, defecation dysfunction, altered sensation of the saddle area and sciatica) measured at 3 different moments (at presentation, FU 1 and FU2), 12 models were created. To correct for possible confounding, the following co-variables were added: gender; age at surgery; duration of CES complaints at presentation; duration of complaints of herniated disc at presentation. Two extra co-variables were added to the models evaluating clinical features at FU1 and FU2: (1) time to decompression and (2) the evaluated clinical feature at presentation (since dysfunction at presentation is correlated with dysfunction at the next follow up moment). Because of anticipated scarce data on sexual dysfunction, sexual dysfunction was not included in nor analyzed by any regression model. In case of quasi-complete separation of data, the concerning variable was not included in the regression model to maintain high quality analysis.

Prior to running regression models, missing values of the following parameters were handled by multiple imputation with five imputation sets: duration of CES complaints at presentation; duration of sciatica at presentation; time to decompression; defecation dysfunction at presentation, at FU 1 and at FU 2; micturition dysfunction at FU 1 and at FU 2; altered sensation of the saddle area at FU 1 and at FU 2; sciatica at FU 1 and at FU 2. Some numerical data were grouped for analyses, e.g. time to decompression was stratified into six groups: <12 hours, 13–24 hours, 25–36 hours; 37–48 hours; 49–72 hours; >72 hours. Two-sided p-values<0.05 were considered statistically significant.

Association between MRI features and clinical presentation

Thirty-seven patients CES (82%) displayed bilateral compression of the cauda equina at MRI, of whom 19 (51%) indicated that their sciatica was unilateral. There was no correlation between MRI and history of the patient for location of sciatica (p = 0.631).

MRI features (severity of cauda equina compression; type of cauda equina compression i.e. unilateral or bilateral; level of disc lesion) were not associated with absence or presence of any of the clinical features (thus micturition dysfunction, defecation dysfunction, altered sensation of the saddle area or sciatica).

A trend was seen for defecation dysfunction at presentation with the co-variable gender, albeit not significant (p = 0.061): women more often suffered from defecation dysfunction at presentation.

Note: the MRI parameter ‘type of compression’ (i.e. uni- or bilateral) was removed from the models evaluating micturition at presentation/ FU 1 and altered sensation of the saddle area at FU 2 due to quasi-complete separation (the parameter type of decompression displayed too little variation when analyzing those outcomes). Because all patients suffered from sciatica at presentation, sciatica at presentation was not analyzed as a predictor for outcome.

Association between MRI features and clinical outcome

MRI features were not demonstrated to be associated with outcome of micturition, defecation, sciatica or altered sensation of the saddle area. The co-variable time to decompression was correlated with sciatica at FU1: a shorter FU time correlated with more sciatica at FU 1 (p = 0.043); this correlation disappeared at FU 2.

Note: Sciatica and altered sensation of the saddle area at presentation were removed from the models evaluating clinical outcome of those functions due to quasi-complete separation ((almost) all patients displayed sciatica and altered sensation of the saddle area at presentation).

Anteroposterior (AP) diameter of the lumbar spinal canal in CES

For 26 CES patients, MRI scans were digitally available and used to measure the AP diameter of the lumbar spinal canal. For comparison, AP diameters of thirty-one patients lumbar herniated disc patients without CES, operated because of sciatica, were measured as well. Patient characteristics known to possibly influence spinal canal size (age, gender) were compared between groups (CES patients with AP measurements; CES patients without AP measurements; lumbar herniated disc patients without CES operated because of sciatica): differences were demonstrated to be non-significant (Table 3). The results of the measurements however, did differ: CES patients displayed a statistically significant smaller lumbar spinal canal diameter at all levels, both disc levels as well as mid-vertebral levels, compared to sciatica patients without CES (largest p = 0.002; Table 4; Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Distribution of the sagittal lumbar spinal canal diameter at vertebral levels (left column) and at disc levels (right column), stratified for CES patients and for sciatica patients.

https://doi.org/10.1371/journal.pone.0186148.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Baseline characteristics of CES patients versus sciatica patients.

https://doi.org/10.1371/journal.pone.0186148.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Measurements of the spinal canal.

The mean sagittal diameter of the spinal canal, measured in millimetres to the nearest tenth. Compared between CES patients and herniated disc patients without CES, operated because of sciatica.

https://doi.org/10.1371/journal.pone.0186148.t004

To compare the measured AP diameters of the CES patients and the sciatica patients with standardized spinal canal diameters reported in literature, studies with a normative distribution of the AP diameter of the lumbar spinal canal, measured at MRI, were searched. Some identified studies were found to be unfit for this comparison since the measured population was biased (e.g. patients referred for low-back pain) [13,14], because the study lacked an exact definition of the subjects for which measurements were taken [15] or because no measurements were available at disc level [16]. The study of Chatha et al. [17] seemed most appropriate for comparison. It describes measurements of the spinal canal in 100 British, symptom-free patients (mean 62 years), who were referred for MRI to screen for presence of metastatic disease, without subsequently having evidence of spinal tumours at the concerning MRI. Even though the study of Chatha et al. is subject to selection bias, the sample size is rather large and patients are quite comparable to the patients in the current study with regard to age, and, in addition, probably quite comparable in terms of race (predominantly Caucasian). In addition, it reports spinal canal size both at intervertebral and disc level, in contrast to aforementioned studies. In order to compare the findings of the current study with the measurements reported by Chatha et al., the average AP spinal canal diameter reported by Chatha et al. was taken as a cut off value. For both the CES patients and the lumbar herniated disc patients without CES, operated because of sciatica, the proportion below the cut off value was indicated (Table 5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Proportion with smaller than average diameter.

The average sagittal diameters that are used as cut off values are the ones reported by Chatha et al.[17]^*.

https://doi.org/10.1371/journal.pone.0186148.t005

Relation to literature

Spinal canal size of CES patients has not been studied before, however, studies about spinal canal size in patients with other spinal diseases are available. Haig et al. for example, compared patients with low back pain, sciatica and lumbar spinal stenosis with controls, concluding that there is no significant difference between patients and healthy subjects with regard to spinal canal measurements [18].

No associations between MRI and clinical features at presentation nor outcome of CES were identified in this study. Since this is the first study to evaluate this correlation, no references are available to state these results. Similar studies have been performed for other spinal diseases such as sciatica [12] or lumbar spinal stenosis [11,19–21], displaying no correlation between imaging and clinical features, being in line with the current study. The suggestion that factors other than the spinal canal size alone, such as local neurovascular problems, venous obstruction or effect of local inflammatory cytokines–factors that are not identifiable at conventional MRI–contribute to differences in clinical manifestation of CES, seems sensible [18]. Gadolinium-Enhanced MRI may be a promising method in this respect: in stenotic dog models, Kobayashi et al. demonstrated local enhancement at the site of the constricted cauda equina at Gadolinium-Enhanced MRI[22].

A non-significant trend was seen between the co-variable gender and defecation dysfunction at presentation, namely: female gender was associated with more defecation dysfunction at presentation (non-significant: p = 0.061). This finding correlates with current literature stating that e.g. constipation is more common in women than in men, in both CES population as well as in the general population [23,24].

The co-variable time to decompression was correlated with sciatica at FU 1: a shorter time to decompression was associated with more sciatica (p = 0.0043), which correlation was not demonstrated for FU 2. This finding does not refute the beneficial effects of early decompression which was demonstrated by others [2,9,10,25–28], but rather indicates a correlation between factors indicating a worse prognosis and shorter time to decompression (guided by clinical decision making), such as acute compression of the cauda equina, which is believed to have a worse prognosis than a more gradual compression [25,29].

Implications

If there truly is a difference in lumbar spinal canal size between lumbar herniated disc patients with CES and lumbar herniated disc patients without CES (operated because of sciatica), this might imply that sciatica patients with a small lumbar canal may need to be approached differently in managing complaints of herniated disc, e.g., it might be beneficial to opt for decompressive surgery earlier in patients with a significant narrow spinal canal size to reduce the risk of developing CES at a later stage. Since this is the only study presently available that evaluated this correlation–and since the setting was retrospective–further prospective research should be conducted before clinical consequences are considered and changes of guidelines are obligatory. A prospective follow up study among sciatica patients would be fitting best, measuring the AP diameters at MRI at presentation—preferably using Gadolinium-Enhanced MRI, also enabling to focus on other factors such as local edema—and ensuring adequate follow up, permitting to correlate the incidence of CES with documented spinal canal size and other MRI features. In case of development of CES, clinical signs and symptoms should be recorded and adequate long term postoperative follow up should take place to evaluate the predictive value of MRI characteristics.

MRI and clinical features were not found to be correlated in the current study. Even though this study has a rather large study population when compared to other CES studies, the limited number of included patients might have caused an inability to detect significant correlations between MRI and clinical features. Aforementioned study proposal with a substantially large cohort and prospective design should be able to give more insights into the predicting value of imaging features in CES patients.

Limitations

This retrospective study design introduces information bias, e.g. complaints might be reported in the file differently than they were meant by the patient, notes are interpreted differently by the researcher than the clinician originally meant, or notes are simply missing. It is impossible to eradicate this bias completely in the current study design, however, the authors believe bias was minimized by careful assessment of medical notes. Multiple imputation was used to deal with missing values, which was believed to be non-problematic due to the assumption of missing at random. The alternative to multiple imputation would be a complete case analysis, which was believed to be more prone to bias [30].

Potential selection bias is introduced with regard to (1) the included CES patients and (2) the samples of CES and sciatica patients for which AP diameter were measured. Firstly, the inclusion criteria of this study correspond to the most used definition of CES. Indeed the broadness of this definition naturally introduces heterogeneity within the studied population. However, this heterogeneity is inherent to CES and is exemplified by the diversity of clinical manifestations. Division into different groups to create more homogeneity per group (e.g., by using the groups of Tandon and Shankaran [31] or the groups CES-retention and CES-incomplete [32]) would be interesting in a larger cohort, preferably with prospective design. Dealing with the presented cohort size and retrospective study design however, substantial risk of improper grouping and thus low quality analysis lures when dividing included patients into different groups.

Secondly, as displayed in the Results section, CES patients for whom AP diameters were taken form a representative sample of the complete CES cohort and are also similar to the sampled sciatica patients without CES in terms of age and gender, parameters known to influence measurements [11,33,34]. Height was not available retrospectively and therefore not included; however, this parameter was described previously as a possible influencer of spinal canal measurements [35] and therefore, height as a confounder cannot be completely eradicated in the current study.

In this study, no information about degree of decompression is available (i.e. no evaluation of MRI scans was done postoperatively). This could introduce some bias in correlating outcome with the MRI features at presentation: in case decompression was less successful, certainly, more complaints will persist at follow up, which might be not related to the initial MRI features at presentation. However, since all patients were decompressed by similar technique in the same centre, variations in decompression were expected to be minimal.

This study used mid-sagittal AP diameter as indicator of spinal canal size instead of area measurements. AP diameter is proven to be well correlated with area measurements [36] and is currently the measurement most often used in studies relating to spinal canal size. The authors thus believe AP diameter to be a reliable indicator of spinal canal size. The quite recently introduced “reduced interlaminar angle” was proven to be a relevant measurement in the stenotic population in particular, however, was seen as less relevant in the current study population [37].