Introduction

Approximately 30% of patients surviving acute carbon monoxide (CO) poisoning display various chronic neuropsychiatric symptoms [31, 32]. Of these, approximately two-thirds demonstrate persistent neurological symptoms from the acute phase to the chronic phase. The remaining one-third show delayed neuropsychiatric sequelae (DNS), which are recurrent neuropsychiatric symptoms occurring after an interval of apparent normality (“lucid interval;” mean duration 22 days) following apparent recovery from acute symptoms [6, 33]. Animal experiments and some clinical studies have led to the hypothesis that damage after CO poisoning results from complicated mechanisms due to CO-mediated toxicity: mitochondrial oxidative stress in the central nervous system (CNS) following CO-induced tissue hypoxia [35]; perivascular oxidative stress mediated by intravascular neutrophil activation [26]; and alteration of myelin basic protein (MBP), a major myelin component in the CNS, due to lipid peroxygenation leading to auto-immunological demyelination of CNS [24, 25]. Auto-immunological demyelination induces further inflammation in the cerebral white matter (CWM) [31]. Gray matter structures, such as the cerebral cortex, basal ganglia and hippocampus, must be damaged by severe hypoxia, since these structures display higher cellular activity and higher oxygen requirements than white matter structures and are more vulnerable to oxygen deprivation [29]. However, damage in the CWM is seen in patients both with and without damage to gray matter structures, and the severity of CWM damage appears to correlate with prognosis in CO-poisoned patients [15, 34].

Assessment of CWM damage caused by CO poisoning in the acute or subacute phase contributes to predictions of progress to DNS and prognosis of chronic symptoms, and appropriate triage of patients with CO poisoning for observation and treatment. Additional quantitative and objective examinations are desirable for assessment of CWM damage after CO poisoning. However, no universally accepted severity scale in routine examinations, such as level of consciousness or carboxyhemoglobin concentration, is available for assessing CWM damage caused by CO poisoning. This is because clinical features are largely affected by the degree of cellular hypoxia resulting from binding of CO to myoglobin rather than hemoglobin and may be markedly affected by various conditions before admission, such as the duration before hospitalization and the care provided before hospitalization [7, 12, 20]. As one of mechanisms for damage in CWM is auto-immunological demyelination, measuring the MBP concentration in the cerebrospinal fluid (CSF) has recently been proposed as an indicator for the extent of CWM damage after CO poisoning [11, 14]. However, detection of MBP using a lumbar tap is a highly invasive procedure and only indicates white-matter damage somewhere within the entire CNS. A less-invasive, objective and quantitative examination that could be used in place of measuring MBP is therefore desired. Diffusion tensor imaging (DTI), a magnetic resonance imaging (MRI) sequence, is potentially more sensitive for detecting demyelination in CWM. Among various quantitative parameters such as apparent diffusion coefficient (ADC) and eigenvalues derived from DTI, fractional anisotropy (FA) has been recognized as the most useful for evaluating the integrity of CWM fibers [2]. Indeed, FA is frequently used for evaluating the extent of damaged CWM fibers in patients with demyelinating diseases such as multiple sclerosis [1, 27]. CO poisoning causes damage in various regions of the CWM, but the centrum semiovale has been considered a region more responsible for chronic neuropsychiatric symptoms after CO poisoning than other regions [4, 10, 19, 22]. Herein, we measured FA from DTI at the centrum semiovale in CO-poisoned patients, and evaluated the correlation between the FA and concentration of MBP in the CSF. This study aimed to clarify whether FA in the centrum semiovale offers a quantitative indicator of the extent of demyelination in damaged CWM during the subacute phase in CO-poisoned patients.

Methods

Patients

All study protocols were approved by the Ethics Committee of Iwate Medical University, Morioka, Japan. Patients recruited to this study were admitted to Iwate Medical University Hospital between April 2008 and February 2011. Entry criteria for this study were: age ≥20 but ≤60 years in patients who had suffered from CO poisoning caused by a fire or charcoal burning; performance of DTI and measurement of MBP concentration according to the protocol in this study; no past history of brain disorders, including surgical operation, irradiation, stroke, infection or demyelinating disease; and provision of written informed consent to participate. Diagnosis was based on present history of exposure to CO and presence of acute neurological symptoms such as impairment of consciousness and headache on admission. After excluding patients who did not meet the entry criteria, 26 patients were enrolled. Mean duration from the scene of CO exposure to arrival at our institute was 5.0 h (range 0.3–81 h). All patients were treated with hyperbaric oxygenation therapy (HBO2) (60 min of 100% oxygen inhalation via mask at 2.8 atmospheres absolute) started within 24 h of admission. HBO2 was continued with a single daily session for a week excluding the weekend. HBO2 was further continued for 4–8 weeks in cases with persistent symptoms. If DNS occurred, HBO2 was restarted and continued until 2 months after CO exposure. HBO2 was discontinued upon patient request or when symptoms were sufficiently improved. Duration of HBO2 administration for all patients ranged from 1 to 60 sessions (mean 12 sessions). The day of CO inhalation was defined as day 1 in this study.

Measurement of MBP concentration in CSF

MBP concentration in the CSF was examined using a lumbar tap at 2 weeks after CO poisoning (between day 12 and day 16) for all patients. Obtained CSF was frozen at −20°C within 1 h after lumbar tap, then the frozen CSF was transported on dry ice to an outside laboratory (SRL, Tokyo, Japan). MBP in the CSF was assayed and measured using a MBP ELISA kit (Cosmic Corp., Tokyo, Japan) immediately after arrival at the laboratory. If the assay was delayed for a long time, frozen CSF was stored at −80°C. An abnormal MBP concentration was defined as ≥102 pg/ml. When the level of MBP was below the limit of detection, the result from the laboratory was reported as MBP ≤40 pg/ml.

DTI

For all patients, DTI was also performed at 2 weeks (between day 12 and day 16) using a 3.0-T whole-body scanner (GE Yokogawa Medical Systems, Tokyo, Japan) and 8-channel coil. Measurements of FA and ADC were performed using data from DTI (repetition time, 10,000 ms; echo time, 62 ms; matrix 128 × 128; field of view, 240 × 240 mm; 4 mm thickness with 1.5 mm gap; 6 motion-probing gradient directions; b value, 1,000 s/mm2). The region of interest (ROI) was manually placed in the bilateral centrums semiovale in the CWM on non-diffusion-weighted images (Fig. 1). FA and ADC were measured bilaterally at the centrum semiovale, using free MRIcro software (http://www.cabiatl.com/mricro/). The FA and ADC for each subject were determined as the mean of values measured twice by the same investigator (S.F.), who was blinded to clinical data. The second measurement was performed 1 week after the first test, using a different randomized order of measurements from the first test. Finally, the mean FA and mean ADC values for the right and left centrum semiovale were calculated and defined as absolute values for each subject. The same procedures described above were performed for 21 age-matched healthy volunteers as controls (18 men, 3 women; mean age 41 ± 10 years, range 22–56 years).

Fig. 1
figure 1

Measurements of FA and ADC value at the centrum semiovale in a patient (case 7 in group S). Regions of interest (ROIs) were placed bilaterally on the centrum semiovale in non-diffusion-weighted image

Full size image

Observation of symptoms

Neurological symptoms were continuously observed for 6 weeks after admission using routine neurological examinations. Patients were assigned to one of two groups according to clinical behavior at 6 weeks (day 40–44) after CO poisoning: group S, patients displaying neuropsychiatric symptoms; group A, patients showing asymptomatic status. Group S included both patients with symptoms persisting for 6 weeks and patients with DNS. DNS was defined as recurrent symptoms after apparent improvement of acute symptoms followed by a lucid interval. General intellectual function was also estimated using the mini-mental state examination (MMSE) [9] at 6 weeks after CO exposure. We defined the normal range, borderline range and dementia according to MMSE scores as ≥27, ≤26 but ≥22, and ≤21, respectively. When scores were considered borderline, patients with educational background ≥9 years and evidence of obvious personality change according to interviews with family members were diagnosed with dementia.

Statistical analyses

We statistically compared differences in mean age among group S, group A and controls using the Mann-Whitney U test. The incidence of abnormal (≥102 pg/ml) MBP concentration was compared between the two patient groups (group S and group A) using the χ 2 for independence test. Mean FA and mean ADC values among two patients groups and controls were compared using the Mann-Whitney U test. Intra-operator reliability for all absolute FA and ADC values was evaluated according to classification of the intra-class correlation coefficient (ICC) [21]. For ICC(1,1) and ICC(1,k) as intra-operator reliability, agreement of all absolute values between the first and second tests was analyzed for right and left lesions using one-factor analysis of variance. After excluding patients showing undetectable concentrations of MBP (≤40 pg/ml), the correlation between MBP and mean FA value was estimated using Spearman’s correlation coefficient by rank test. Statistical significance was established at the p < 0.05 level in all analyses.

Results

A total of 51 patients were admitted to our institute for treatment of CO poisoning between April 2008 and February 2011. After excluding 25 patients who did not meet the entry criteria for this study, a total of 26 patients (24 men, 2 women; mean age 40.1 ± 11.4 years) were enrolled. All patient data are summarized in Table 1. In 19 (73%) of 26 patients, acute symptoms resolved completely within 4 days after admission, and no neuropsychiatric symptoms were present at 6 weeks from CO-inhalation (group A). The remaining seven patients (27%) displayed chronic neuropsychiatric symptoms at 6 weeks (group S), including four patients with continuous persistence of symptoms for 6 weeks and three patients exhibiting DNS after apparent improvement of acute symptoms followed by a lucid interval. DNS in three patients occurred after DTI and measurement of MBP on day 21 in case 5, day 19 in case 6 and day 18 in case 7. Mean age was 45 ± 12 years in group S, 38 ± 11 years in group A and 41 ± 10 years in controls. No significant differences in age were found between groups S and A (p = 0.24), between group S and controls (p = 0.51), or between group A and controls (p = 0.40).

Table 1 Summary of all patients
Full size table

In the seven patients in sroup S, six showed abnormal MBP concentrations (≥102 pg/ml), and one patient showed a level of 99 pg/ml. None of the 19 patients in group A showed abnormal concentrations of MBP, with 16 patients showing undetectable concentrations of MBP (≤40 pg/ml). The incidence of an abnormal MBP levels was statistically different between groups P and A (p < 0.001). MBP concentrations for the four patients with persistent symptoms in group S, for the three patients with DNS in group S and for the three patients in group A were more than ≥150 pg/ml, around 100 pg/ml and around 50 pg/ml, respectively (Table 1).

Table 2 shows ranges and means of FA and ADC for each group. The range of FA for group S slightly overlapped that for group A, but differed markedly from that for controls. Ranges of FA for group A and controls were similar. The mean FA for group S was significantly lower than those for group A (p < 0.001) and controls (p < 0.001), whereas no significant difference was found between group A and controls (p = 0.57) (Fig. 2a). In Fig. 2a, individual mean FA values of the three patients with DNS were not obviously different from those of the four patients with persistent symptoms in group S. Group S patients were clearly differentiated from group A patients at a cutoff of 0.353 (100% sensitivity, 94.7% specificity) and from controls at a cutoff of 0.360 (100% sensitivity, 100% specificity). On the other hand, the range of ADC in each group was similar, and the mean ADC did not differ significantly among any of the three groups (Fig. 2b). Intra-operator reliability for absolute FA was classified as “almost perfect” for the centrum semiovale bilaterally; ICC(1,1) and ICC(1,k) were 0.88 and 0.93 for the right side, and 0.95 and 0.98 for the left side, respectively. Intra-operator reliability for absolute ADC was also classified as "almost perfect" for bilateral centrum semiovale; ICC(1,1) and ICC(1,k) were 0.98 and 0.99 for the right side, and 0.91 and 0.95 for the left side, respectively.

Table 2 Range and mean value of FA and ADC for each group
Full size table
Fig. 2
figure 2

Differences of mean FA (a) and mean ADC (b) values in the centrum semiovale bilaterally among group S, group A and controls. In group S, black and white squares represent patients with DNS and persistent symptoms, respectively (*p < 0.001).

Full size image

After excluding 16 patients showing undetectable levels (≤40 pg/ml), the ten remaining patients (all patients in groups S and 3 patients in group A) showed a strong correlation between the mean FA and MBP (r = −0.79, p = 0.02) (Fig. 3).

Fig. 3
figure 3

Correlation between FA and MBP concentration in ten patients showing MBP concentration >40 pg/ml. White circle, patient with persistent chronic symptoms in group S; black circle, patient with DNS in group S; triangle, patient in group A

Full size image

Discussion

Ide et al. [11] have documented that MBP concentration in patients with DNS showed marked elevation around 2 weeks after CO poisoning, peaking at around 30 days. The timing for MBP measurements in the present study was thus established at 2 weeks (between day 12 and day 16) after admission. As a result, the incidence of abnormal MBP concentration was significantly higher in patients with chronic neuropsychiatry symptoms (group S) than in patients without chronic symptoms (group A) or controls. These results suggest that patients in group S certainly suffered from demyelinating changes somewhere in the CWM and support the theory that chronic neuropsychiatric symptoms after CO intoxication result from progressive demyelination in the CWM [24, 25]. The MBP concentration in case 6 was slightly lower (99 pg/ml) than the abnormal level, but the patient displayed akinetic mutism compatible with DNS at 1 week after measurement of MBP. The concentration of MBP in this patient might have been on the way to reaching abnormal levels, as demyelination of CWM in patients with DNS has been considered to undergo gradual progression during the lucid interval [13]. MBP concentrations of DNS patients were between those of patients with persistent symptoms in group S and those of patients in group A (Table 1; Fig. 3). These findings may indicate that demyelination begins to progress during the lucid interval before DNS. Although measuring MBP concentrations thus offers a useful indicator for assessing the extent of demyelination due to CO poisoning, detection of MBP using a lumbar tap is a highly invasive procedure and only indicates white-matter damage somewhere within the CNS.

Neuroimaging is minimally invasive and can visualize any region in the CWM. T2-weighted imaging (T2WI) often depicts abnormalities in the CWM in CO-poisoned patients. However, the interpretation of findings from routine MRI is difficult, as hyperintense foci in the CWM on T2WI can represent various progressive histological changes, including vasogenic edema, multiple necrosis, extensive axonal destruction and/or demyelination without axonal destruction [4, 12]. We therefore performed DTI in the same period as detection of MBP, since DTI is potentially more sensitive for assessing the extent of demyelinating changes in the CWM than other MRI sequences. As progressive reduction of FA values with age has been reported [5], we compared patients <60 years old with age-matched controls in this study. The finding of no significant difference in mean age among groups S, group A and controls suggests a negligible contribution of aging to FA values in this study. Previous reports have documented damage in various regions of the CWM after CO poisoning [8, 18, 28]. Indeed, some studies have reported correlations between FA values in various regions of the CWM in the chronic phase and cognitive dysfunction among CO-poisoned patients with DNS [16, 23, 30]. However, the centrum semiovale in the CWM has been suggested as a key region responsible for chronic neurological symptoms [4, 10, 19, 22]. A study using DTI at various phases after CO poisoning has also shown that FA in the centrum semiovale changes in parallel with cognitive impairments or neurological symptoms [17]. Based on these reports, we placed the ROI on the centrum semiovale to measure FA and ADC from DTI. As a result, mean FA for group S presenting with chronic neuropsychiatric symptoms was significantly lower than that for group A presenting with no chronic symptoms or that for controls consisting of healthy volunteers, whereas no significant difference was evident between group A and controls. In contrast, mean ADC did not differ significantly among the three groups. FA must be more sensitive for detecting CWM damage than ADC. Furthermore, these findings suggest that white matter fibers in the centrum semiovale were demyelinated in the subacute phase (2 weeks after poisoning) in CO-poisoned patients presenting with chronic symptoms. Notably, reductions in FA, suggestive of demyelination, were already present in the centrum semiovale before the recurrence of symptoms in the three patients with DNS. The reliability of this finding is supported by the result that MBP concentrations in DNS patients showed greater increases than those in group A patients at 2 weeks. These findings indicate the possibility of using FA in the centrum semiovale as an appropriate examination for predicting DNS during the lucid interval.

Our pilot study of DTI for CO-poisoned patients showed that FA enables representation of damage to white matter fibers in the centrum semiovale of patients with chronic neuropsychiatric symptoms [3]. That report, however, failed to demonstrate any correlation between FA in the centrum semiovale and MBP concentration, presumably because of the small sample size. Although subject criteria were more strictly established in the present study than in our previous investigation, the greater number of subjects in this study allow us to show a linear correlation between FA and MBP in ten patients showing MBP concentrations >40 pg/ml. This finding validated the use of the centrum semiovale to represent various demyelinated lesions in the CWM, and FA in the centrum semiovale obviously offers a quantitative indicator of demyelination in CO-poisoned patients with chronic neuropsychiatric symptoms.

Some limitations must be considered in the interpretation of the study results. First, FA in the centrum semiovale may not strictly mirror the amount of demyelination in the whole CWM, although FA in the centrum semiovale correlated with MBP concentration. In group S, FA values in the centrum semiovale of the three DNS patients were not clearly different from those of the four patients with persistent symptoms (Fig. 2a), whereas MBP seemed to allow differentiation between subgroups in group S (around 100 pg/ml in patients with DNS and ≥150 pg/ml in patients with persistent chronic symptoms). This discrepancy might hypothetically be explained if demyelinated lesions in patients with persistent symptoms vary more than those in DNS patients. FA measured in this study suggests the magnitude of demyelination in the centrum semiovale, whereas MBP concentration not only indicates the magnitude, but also the width of demyelination in the whole CNS. We think that FA in the centrum semiovale cannot allow differentiation of the severity of CWM damage among subjects including patients with DNS and those with persistent symptoms. Second, the chronic neuropsychiatric symptoms seen after CO poisoning may not be solely attributable to demyelinating changes in fibers of the centrum semiovale. However, knowing to the focus on the region of the CWM is obviously very useful when evaluating the extent of CO-induced CWM damage using neuroimaging. We considered that the centrum semiovale represents the main region of damage and should be the focus of attention on neuroimaging in the subacute phase after CO poisoning [10]. Third, the sample size in this study was still small, with markedly fewer subjects in group S than in group A. The small number of DNS patients resulted in difficulties with statistical comparisons between subgroups in group S and other groups. However, the small sample size resulted from the strict entry criteria for this study. Furthermore, we did not select subjects with any bias other than the criteria established for this study. Indeed, percentages for patients with and without chronic symptoms in this study were in agreement with the results of previous reports [6, 33]. Fourth, findings in this study cannot be applied to patients over 60 years old. In senior patients, FA values may be overestimated as aging may lead to reduced FA values.

Conclusions

This is the first report to find that FA in the CWM correlates with MBP concentrations in the CSF during the subacute phase in CO-poisoned patients. The identification of a significant negative correlation between FA in the centrum semiovale and MBP concentration validates the concept that the centrum semiovale can reveal various demyelinated lesions in the CWM and that FA in the centrum semiovale offers a quantitative indicator of demyelination in CO-poisoned patients with chronic neuropsychiatric symptoms.