Introduction

Diagnosis of thoracic myelopathy tends to be delayed because of its rarity [1, 2] and its diverse neurological manifestations. These diverse manifestations result from the anatomical uniqueness of the thoracic spine. The neurological symptoms resulting from thoracic disorders are sometimes misdiagnosed as cervical or lumbar disorders [3]. Misdiagnosis can lead to a prolonged preoperative disease duration, which can cause irreversible neurological damage and even unnecessary surgical procedures performed on the wrong spinal region [4]. Although previous studies report the clinical outcomes of different surgical procedures or the radiographic risk factors for poor clinical outcomes [5–9], possible correlations between preoperative manifestations and anatomical pathology [i.e., ossification of the posterior longitudinal ligament (OPLL), ossification of the ligamentum flavum (OLF), intervertebral disc herniation (IDH), and spondylosis (SPD)] or compressed segments (most severely involved anatomical segments) have not been discussed. In the present study, we retrospectively examined 205 patients with thoracic myelopathy to identify symptoms that were specific for anatomical pathology or compressed segments.

Materials and methods

The patient data were collected through retrospective review of charts and radiographic images at eight affiliate institutes of Osaka University. We identified a total of 205 patients who underwent surgery for degenerative thoracic myelopathy from 2000 to 2011. The disease distribution included OLF in 106 patients, OPLL in 17, OLF with OPLL in 17, IDH in 23, OLF with IDH in 3, and SPD in 39 (Table 1). There were 143 men and 62 women, and their mean age was 62.2 years (range 21–87 years). The mean preoperative disease duration was 595 days (range 6–6,448 days). The initial complaints, preoperative complaints, compressed segments, and preoperative disease duration were assessed. In addition, we assessed the clinical examination findings, including deep tendon reflexes (DTRs), ankle clonus, Babinski sign, lower limb muscle strength, and the JOA score (full score, 11 points) [10]. In the present manuscript, the initial complaints are indicated by “complaints (onset)” and preoperative complaints are indicated by “complaints (pre-op).” Comorbidities that can affect neurological evaluation included diabetes, which was present in 42 patients (20 %).

Table 1 Patient demographics
Full size table

The complaints (onset) and complaints (pre-op) were further classified into 17 categories, which are listed in Fig. 1. For example, low back pain at onset was labeled low back pain (onset), and low back pain occurring immediately before surgery was labeled low back pain (pre-op). Complaints presented without “(onset)” or “(pre-op)” were used to describe all complaints, regardless of their timing.

Fig. 1
figure 1

Frequency of complaints (onset, pre-op). The most frequent complaint was lower limb numbness (onset 50 %, pre-op 78 %), followed by subjective lower limb weakness (onset 28 %, pre-op 67 %) and unsteady gait (onset 17 %, pre-op 41 %). Low back pain indicated lumbar back pain, middle back pain indicated thoracic back pain, and trunk radicular pain indicated radicular pain in the trunk after excluding lumbar and thoracic back pain

Full size image

The DTRs evaluated for the upper limbs included Hoffmann’s sign, Wartenberg’s sign, and those for the biceps, brachioradialis, and triceps tendons. The lower limb DTRs evaluated included the patellar tendon reflex (PTR) and Achilles tendon reflex (ATR). Pathological reflexes, such as ankle clonus and the Babinski sign, also were examined.

Deep tendon reflexes were classified into three categories: hyperreflexia, normoreflexia, and hyporeflexia. Hyperreflexia was defined as having an exaggerated DTR in either or both limbs. Ankle clonus and the Babinski sign also were categorized as positive or negative. Preoperative lower limb muscle strength included assessment of the iliopsoas (IP), quadriceps femoris (QF), tibialis anterior (TA), and gastrocnemius (GS) using manual muscle testing (MMT 0–5 grades). Manual muscle testing grades <4 were defined as clinically weak.

Bladder dysfunction was evaluated based on the JOA bladder subscore (0–3) [10]. A grade of zero or one was considered severe dysfunction.

The compressed segments were defined as the segments that required operation, which were confirmed with T2-weighted magnetic resonance imaging (MRI) in 202 patients or computed tomography after myelography (CTM) in 3. Cervical and lumbar MRI or CTM were performed simultaneously with thoracic imaging to exclude coexistence of cervical or lumbar diseases. We determined the level at which the conus medullaris (TLCM) terminated in 115 (80 %) of the 144 patients who had compression from T10/11 to L1/2. The termination level was defined with reference to the vertebra or intervertebral disc (upper, middle, lower vertebra level, or intervertebral disc level).

Because the distributions of preoperative disease duration was bimodal (except for those with duration >600 days) the patients were assigned to either a short-duration group (<180 days; 91 patients) or a long-duration group (≥180 days and <600 days; 62 patients; Fig. 2).

Fig. 2
figure 2

Distribution of preoperative duration. The histogram shows the distribution of preoperative disease duration. A bimodal distribution was evident after excluding patients with preoperative disease duration longer than 600 days

Full size image

PASW Statistics software version 18 (SPSS Inc., Chicago, IL) was used for statistical analysis. Comparisons between two groups were performed with the Mann–Whitney U test and Fisher’s exact test. In these comparisons, p values <0.05 were considered significant. Multivariate logistic regression analyses with forward stepwise selection were performed to assess predictors of preoperative manifestations. Before performing the multivariate logistic regression analyses, we confirmed that correlation coefficients between any two independent variables were below 0.7. Comparisons among four groups were performed with the Mann–Whitney U test after the Kruskal–Wallis test; a corrected p < 0.0083 (corrected by the Bonferroni method) was used to indicate significance.

Results

The most frequent complaint was lower limb numbness (onset 50 %, pre-op 78 %), followed by subjective lower limb weakness (onset 28 %, pre-op 67 %), and unsteady gait (onset 17 %, pre-op 41 %; Fig. 1). The frequency of all complaints increased from onset to surgery. Lower limb muscle weakness was observed in 67 (33 %), 35 (17 %), 60 (29 %), and 39 (19 %) patients for the IP, QF, TA, and GS, respectively. The distribution of the preoperative urinary function scores were grades 0 (12, 6 %), 1 (40, 20 %), 2 (74, 36 %), and 3 (79, 39 %). The average preoperative JOA score was 5.4 (range 0–10.5). After excluding 76 patients with upper limb hyperreflexia or diabetes mellitus, we examined the PTR, ATR, ankle clonus and the Babinski sign in 129 patients (Table 2).

Table 2 Distribution of PTR, ATR, ankle clonus, and Babinski sign
Full size table

Disease distribution in compressed segments and comparison of patient demographics according to anatomical pathology

Among the patients with OLF, the distribution was predominantly at the T9/10, T10/11, and T11/12 segments in 70 % of all compressed segments. Among those patients with OPLL, the distribution was mainly in the upper or middle thoracic segments. Intervertebral disc herniation and SPD were seen at the lower thoracic segments (Fig. 3).

Fig. 3
figure 3

Distribution of compressed segments based on anatomical pathology

Full size image

Patients who had a combination of two diseases were excluded in the comparison of diseases. Compared to OLF, IDH, and SPD, OPLL was more frequent in women than men. Intervertebral disc herniation had shorter preoperative disease duration than OLF an OPLL, and those with OPLL had lower preoperative JOA scores than those with OLF and IDH (Table 1).

Predictors for preoperative manifestations

To assess the predictors for each preoperative manifestation, multivariate logistic regression analyses with forward stepwise selection were applied to the potential independent variables, including anatomical pathology, compressed segments, age, sex, diabetes mellitus, and upper limb hyperreflexia (Table 3). All four anatomical pathology variables showed an uneven spatial distribution of compressed segments in terms of the rostral–caudal and anterior–posterior directions. To account for the possibility that anatomical pathology and compressed segments were confounding variables, we performed the analyses in two steps. In the first step, anatomical pathology and compressed segments were included as predictors without distinction between anterior and posterior directions. In the second step, after excluding anatomical pathology, compressed segments were subdivided based on anterior or posterior compression. If compressed segments were significant predictors at the second step, then compressed segments should be significant rather than anatomical pathology. On the other hand, if the second step analysis showed no significant compressed segments despite the identification of significant anatomical pathology in the first step, then anatomical pathology was considered a significant predictor. The compressed segments were divided into four categories that included C7/T1-T9/10, T10/11, T11/12, and T12/L1. On multivariate analysis, six preoperative manifestations were significant predictors with odds ratios greater than three and a p < 0.05, except for diabetes mellitus and upper limb hyperreflexia. Strong positive correlations were found among the following factors: (1) low back pain and IDH or T11/12 compression; (2) foot drop and IDH; (3) hyporeflexia in the PTR and IDH; and (4) weak QF, TA or GS and IDH.

Table 3 Multivariate logistic regression analysis including anatomical pathology
Full size table

In addition, strong negative correlations (odds ratio <0.33 and p < 0.05) were found between: (1) trunk numbness and T11/12 compression; (2) lower limb pain and C7/T1-T9/10 compression; (3) unsteady gait/Babinski sign and T12/L1 compression. Age was a significant positive predictor for hyporeflexia in the PTR and ATR, and a significant negative predictor for ankle clonus.

Differences based on anterior–posterior compression

As the next step, compressed segments were subdivided based on anterior or posterior compression, independent of the anatomical pathology to avoid confounds between these two variables. The independent variables were compressed segments (divided between anterior and posterior at C7/T1-T9/10, T10/11, T11/12, and T12/L1), age, sex, diabetes mellitus, and upper limb hyperreflexia (Table 4; Fig. 4). Other than diabetes mellitus and upper limb hyperreflexia, there were strong positive correlations (odds ratio >3 and p < 0.05) between: (1) low back pain and T11/12 anterior compression; (2) foot drop/hyporeflexia in the PTR and T12/L1 anterior compression; and (3) weak QF, TA, or GS and T10/11 anterior compression. Strong negative correlations (odds ratio <0.33 and p < 0.05) were found between: (1) lower back pain and posterior T10/11 compression; (2) trunk numbness and posterior T11/12 compression; and (3) lower limb pain and posterior C7/T1-T9/10 compression.

Table 4 Multivariate analyses by anterior–posterior compression excluding anatomical pathology
Full size table
Fig. 4
figure 4

Schematic showing significant correlations between compressed segments and the manifestations

Full size image

Comparisons between short- and long-duration groups

The preoperative JOA score was significantly lower in the short-duration group than in the long-duration group. The most frequent initial complaint by patients in the short-duration group was subjective lower limb weakness, which was less frequent in the long-duration group. “Abnormal tightness in the lower limbs” (onset) was seen only in the long-duration group. The incidence of weak IP, QF, and GS muscles was higher in the short-duration group than in the long-duration group (Table 5).

Table 5 Comparison between short-duration and long-duration groups
Full size table

Of the 113 patients who had compression at a single segment, patients with T11/12 compression (n = 43) had a significantly longer preoperative disease duration [1,063 days (range 16–5,933 days); p = 0.004] than the other patients [n = 70, 341 days (range 6–2,128 days)].

Termination level of the conus medullaris

The TLCM was from the middle-third of the L1 vertebra to the L1/2 intervertebral disc in 65 % of patients. There was no significant difference for sex. Correlations between the TLCM and the aforementioned level-dependent manifestations, including lower limb muscle weakness (T10/11 anterior compression), lower limb pain (T11/12 anterior compression), low back pain (T11/12 compression), and hyporeflexia in the PTR/foot drop (T12/L1 anterior compression) were investigated. Only low back pain had a significant relationship with TLCM (Fisher’s exact test, Table 6). In the majority of patients with low back pain concomitant with T11/12 compression, TLCM was at the L1/2 intervertebral disc level and the upper-third of the L2 vertebra.

Table 6 Frequency distribution of the conus medullaris termination level
Full size table

Discussion

Neurological manifestations caused by lower thoracic and thoracolumbar lesions are often classified as the epiconus syndrome, conus medullaris syndrome, or cauda equina syndrome [11]. However, the clinical manifestations of these lesions are diverse and complicated to diagnose because of the anatomical complexity of thoracic region [3, 4, 12–14]. The present study showed that some neurological manifestations of lower thoracic or thoracolumbar lesions had strong correlations with the specific anatomical pathology and compressed segments.

Relationships between the manifestations and anatomical pathology

Among the anatomical pathologies, IDH more frequently showed correlations with the manifestations. Intervertebral disc herniation was a significant predictor for low back pain, foot drop, PTR hyporeflexia, and lower limb weakness. On the other hand, OPLL emerged as a significant positive predictor for severe urinary dysfunction because no significant compressed segments were identified for severe urinary dysfunction in the second step of the multivariate analysis.

Relationships between manifestations and compressed segments by anterior and posterior compression

Patients with low back pain were predominantly compressed at T11/12, regardless of anterior or posterior compression. One explanation for this relationship is the involvement of upper lumbar nerve roots including the L2 nerve root. Although there are individual differences in the TLCM [15] (TLCM in patients who showed low back pain and T11/12 compression were largely localized at the L1/2 intervertebral disc level and the upper-third of the L2 vertebra), the T12 to L3 nerve roots flank the spinal cord at the T11/12 level (Fig. 5) [16]. Involvement of the L2 nerve root, whose infiltration is proposed to be an effective treatment for low back pain [17, 18] may explain why patients with T11/12 compression showed a high prevalence of low back pain. Patients with T11/12 anterior compression showed a higher rate of lower limb pain, and there was a significant correlation between T12/L1 anterior compression and foot drop. These correlations also may be explained by the involvement of lumbar nerve roots [16]. T12/L1 anterior compression was implicated as preoperative predictors for hyporeflexia in the PTR, which can be related to lumbar nerve root involvement (Fig. 5). Patients who had T10/11 anterior compression were more likely to have lower limb muscle weakness. Given the fact that only T11 and T12 nerve roots exist at this level (Fig. 5) [16], this muscle weakness is explained by long tract involvement or the involvement of spinal cord segments L2 to L5 if they were located more rostally than previously thought [11].

Fig. 5
figure 5

Schematic showing significant correlations between compressed segments and the manifestations. a T10/11. The T11 and T12 on the top are dorsal sensory roots and the T11 and T12 on the bottom are ventral motor roots. b T11/12. The roots on the top are sensory and the roots on the bottom are motor. c T12/L1. The upper neural elements are sensory and the lower elements are motor

Full size image

Common non-specific manifestations

Among the frequent complaints of patients with thoracic myelopathy, compressed segments and anatomical pathology were not identified as positive predictors for lower limb numbness or unsteady gait. However, only T11/T12 compression was negatively associated with subjective lower limb weakness, meaning it had a low probability if patients showed subjective weakness. As a previous report indicated [19], gait disturbance is one characteristic symptom of myelopathy. The analysis of gait may clarify the characteristics of thoracic myelopathy, and may be valuable in the level diagnosis. Unfortunately, that analysis was not included in the present study, but further work including information related to gait disturbance is planned.

Factors associated with longer preoperative duration

Previous reports indicated that a longer preoperative duration is a significant predictor for poor results after surgery for thoracic myelopathy [6, 20–22]. The present study showed that elusive complaints such as “abnormal tightness in the lower limbs” are related to a longer preoperative duration. In contrast, serious manifestations that impair activities of daily living, such as subjective lower limb weakness and objective lower limb muscle weakness, can significantly shorten the preoperative duration. These manifestations can lead doctors to examine the thoracic region. T11/12 compression, which correlates with low back pain, also correlates with long preoperative duration. Such patients may be treated for lumbar disease because T11/12 is not often included in lumbar MRI.

This study has several limitations. We cannot conclude that non-specific symptoms such as lower limb numbness and unsteady gait were caused by thoracic myelopathy because we did not collect postoperative data on the corresponding symptoms. Furthermore, it is difficult to show direct correlations between manifestations and nervous system anatomy, such as nerve root locations and spinal cord segments. However, our study can add new information to the anatomical distribution of nerve roots and spinal cord segments in lower thoracic lesions.

Conclusion

This multicenter study clarified statistical correlations between compressed segments and the complex clinical manifestations caused by thoracic myelopathy. These correlations will help to better direct our attention to thoracic diseases and may facilitate early diagnosis though additional imaging investigations such as MRI or CTM, which are necessary for definitive diagnosis.