The term “toxic leukoencephalopathy” encompasses a wide spectrum of diseases that may injure and cause structural alteration of the white matter; the insults may be toxic metabolic secondary to chemotherapy or immunosuppressive therapy, environmental, or infectious in origin [1]. On early imaging of some leukoencephalopathies, diffusion-weighted imaging (DWI) occasionally shows a symmetric decrease in the apparent diffusion coefficient (ADC) values in the corona radiata (periventricular) and the centrum semiovale (supraventricular) white matter (which we collectively refer to herein as “periventricular white matter”) [1-19]. In the acute or subacute phase, these insults can be relatively symmetric and involve the periventricular white matter on DWI out of proportion to their appearance on FLAIR images. There are limited reports of the reversibility of this reduced diffusion in some disorders on follow-up MRI studies of toxic leukoencephalopathy [1-16]. Causes that have been described as bringing about potentially reversible toxic leukoencephalopathy include chemotherapy, immunosuppressive therapy, antimicrobial medications (e.g., metronidazole), environmental toxins (e.g., carbon monoxide), and drug abuse [1-19]. Other insults that are not typically classified as toxic leukoencephalopathy may show a similar distribution of reduced diffusion, such as subacute hypoxic-ischemic encephalopathy (HIE) and congenital genetic disorders; these entities usually result in varying degrees of neurologic sequelae [1, 15-21].

It is important to distinguish the MRI appearance of acute toxic leukoencephalopathy from another reversible entity, posterior reversible encephalopathy syndrome, which may be caused by many of the same medications or drugs; however, posterior reversible encephalopathy syndrome is thought to involve the cortex and subcortical white matter in early cases on FLAIR and to extend into the periventricular white matter only in severe cases, with reduced diffusion in only a minority [22, 23]. Of note, radiation damage can be another cause of periventricular white matter injury with FLAIR hyperintensity from small-vessel ischemic demyelination, but this only rarely presents acutely in the periventricular white matter and typically lacks confluent, diffuse involvement with acutely reduced diffusion; rather, radiation injury typically induces focal or multifocal changes with progressively increasing diffusion and decreasing anisotropy in the acute and subacute phases after exposure [24].

This study describes our experience with a group of patients with leukoencephalopathy secondary to a variety of toxic metabolic insults who presented with acute neurologic deficits and a common MRI pattern of symmetric decreased diffusion in the periventricular white matter. The purpose of this article is not only to describe a common and reversible imaging appearance on DWI among various causes in the early stage of this disorder in order to augment recognition and treatment but also to determine whether the acute or subacute MRI findings have prognostic value by correlating the initial MRI markers with follow-up MRI and with the clinical outcome. Nontoxic causes that can mimic this appearance are also described.

This study was approved by the institutional review boards of two hospitals, one of which is a level 1 trauma referral center and the other a tertiary care center with a large number of referrals for oncologic therapy and organ transplantation. Cases of symmetric reduced diffusion in the corona radiata (periventricular) and centrum semiovale (supraventricular) white matter (referred to herein as “periventricular white matter”) in patients with an acute neurologic presentation had been saved in a file through the years 1999-2007. Thirtynine such patients were accumulated during this period and were evaluated retrospectively. Their MRI studies were reviewed and graded jointly via consensus of two staff neuroradiologists. The clinical records and laboratory data were also extensively reviewed for each patient.

Over the 8-year period of this review, 32 patients (17 male, 15 female; mean age, 35.8 years; age range, 11-70 years) presented with an appearance of bilateral symmetric periventricular white matter reduced diffusion on initial MRI that was found to be related to a toxic leukoencephalopathy, as documented in Figures 1, 2A, 2B, 2C, 2D, 3A, 3B, 3C, 4A, 4B, 4C, 4D, 4E, 4F, 5A, 5B, 5C, 5D, 5E, 5F, 6A, 6B, 6C, 6D, 6E, and 6F. Table 1 provides the symptoms, locations, suspected causes, imaging grades, and the clinical outcomes regarding these patients with toxic leukoencephalopathy. In six of these 32 patients, a causative agent was not definitively identified; however, four of the six had received chemotherapy or immunosuppressive medications and hence are listed under the heading “Medication-related” in Figure 1. As in Figure 1 and Table 1, the most common general classification causing acute toxic leukoencephalopathy was medication-related (n = 19), whether from immunosuppressants, chemotherapy, or antimicrobials. Also, 13 of the 32 patients had a history of either solid organ or bone marrow transplantation (denoted by “a” next to cause in Table 1), which is notable because five of the 10 patients with poor outcomes were posttransplantation.

Of the 39 patients initially presenting with symmetric periventricular white matter reduced diffusion on MRI and acute neurologic symptoms, one patient, who had fludarabine toxicity, was excluded from the statistical analysis because of our inability to generate ADC values. Six other patients who presented acutely or subacutely with neurologic symptoms were excluded from the study because of their having nontoxic causes of decreased diffusion in the periventricular white matter (Table 2). Three of these patients had HIE and presented with mild cortical hyperintensity on FLAIR—a finding not present in the patients with toxic leukoencephalopathy—and periventricular white matter reduced diffusion; these three were excluded from the statistical analysis because of the nontoxic cause of their illness (Figs. 7A, 7B, and 7C). Notably, each patient with HIE had a > 70% reduction in ADC and a poor clinical outcome (coma or death). In addition, three children who presented acutely with congenital genetic disorders (all leukodystrophy) showed bright signal centrally in the periventricular white matter on DWI without ADC reduction, consistent with an advancing edge of demyelination (Figs. 8A, 8B, and 8C). The region of reduced diffusion that was clearly visible on the ADC maps in these three patients was predominantly along the periphery of the area of demyelination, a finding that permitted distinction on MRI of these genetic causes from toxic leukoencephalopathy. These three children were also excluded from the statistical analysis because of the nontoxic cause of their illness. Of these six excluded patients, three underwent long-term (> 30 days) repeat MRI, each of which confirmed a lack of improvement on FLAIR MRI.

The ADC values for the entire cohort of 32 cases of potentially reversible acute toxic leukoencephalopathy are listed in Table 3, with the percentage of decrease measured in the affected periventricular white matter relative to the unaffected periventricular white matter; the time to follow-up MRI is also provided. Fourteen of the 32 patients with acute toxic leukoencephalopathy had follow-up MRI and DWI in the chronic phase ≥ 21 days after initial presentation; however, in one of these the ADC values could not be calculated at follow-up (Tables 1 and 3). In these 13 patients, a significant difference was seen between the ADC values in the affected periventricular white matter versus those in the unaffected periventricular white matter on the initial MRI (p < 0.0001, paired Student's t test). In all but one of the 13 patients, the periventricular white matter abnormalities on DWI resolved (while remaining variably present on repeat FLAIR imaging), as evidenced by the lack of a significant difference in ADC values (p = 0.13, paired Student's t test) between those same initially abnormal areas on the repeat MRI examinations as compared with the areas of unaffected periventricular white matter, and as evidenced by the significant difference in affected white matter on follow-up versus initially (p = 0.0014) in those nine who were imaged on the same magnet. The only patient in whom the diffusion abnormality did not entirely resolve had a severe insult (based on FLAIR and DWI) after opiate inhalation; eventually the reduced diffusion resolved, and the patient's symptoms had mostly improved several months later. Additionally, no significant correlation was seen between the initial extent of reduced diffusion and the extent on repeat ADC/DWI (r = 0.19, p = 0.514) in these 13 patients, also confirming that these abnormalities had also visually resolved or nearly resolved on ADC/DWI in all but one patient.

The various correlations measured in this study are depicted in Table 4, which includes all 32 patients with acute toxic leukoencephalopathy. None of the initial MRI markers (percentage of ADC decrease, FLAIR extent, or DWI extent) had significant correlation with the clinical outcome. The only significant correlation was a moderate correlation of the initial extent of periventricular white matter FLAIR abnormality (r = 0.441, p = 0.047) with the residual extent of abnormality on the follow-up FLAIR examination. Figure 9 plots the percentage of ADC reduction in each outcome category, illustrating that the clinical outcome appeared overall unrelated to the percentage of reduction in ADC on the initial MRI.

Of the 32 patients with toxic leukoencephalopathy, 21 underwent follow-up MRI; of those, the FLAIR extent improved in seven. Four of these seven had follow-up MRI at ≥ 120 days after presentation; two developed mild atrophy and the other two did not develop any significant atrophy. Of all 10 patients with follow-up MRI examinations at ≥ 120 days, five did not develop significant atrophy, four developed mild atrophy, and one developed moderate atrophy. None of the 29 patients who received IV contrast material exhibited significant contrast enhancement on the initial examination.

Histologic examination of the involved white matter was available in three patients with severe involvement by toxic leukoencephalopathy (Figs. 3A, 3B, and 3C). In two patients who died, autopsy revealed white matter pallor, neuropil vacuolation, axonal swelling with spheroids, oligodendroglial swelling, enlarged spaces in or between myelin sheaths and axons, and diffuse macrophage infiltration in the white matter without necrosis. In the third patient with available histology, biopsy showed similar findings with macrophage infiltration and abundant axonal spheroids as well as progression to necrosis in areas of abundant macrophage infiltration. This was the only patient in this series to show contrast enhancement (not present initially but later present peripherally) in the involved white matter. Autopsies were refused by the families of the five other patients who died.

Brief mentions follow regarding the major findings in the 13 patients with an acute presentation of toxic leukoencephalopathy that were not related to a known medication. In the two patients with acute liver failure of unknown cause, both had severe hyperammonemia and reduced diffusion in the thalami and posterior limbs of the internal capsules (Figs. 4A, 4B, 4C, 4D, 4E, and 4F); the symptoms resolved after treatment, with resolution of abnormal liver function and hyperammonemia. In a hypertensive patient, there was reduced diffusion in the periventricular white matter with improvement after antihypertensive therapy, but the symptoms recurred 1 month later; repeat MRI at that time showed mild cortical edema typical of posterior reversible encephalopathy syndrome and progression of the DWI abnormalities that subsequently resolved with therapy. A patient with prolonged hyperglycemia had been diagnosed 1 month earlier with Cushing's disease; various therapies led to clinical improvement, although the cause of the hyperglycemia was not ascertained. Each of the three patients with illicit drug inhalation clinically improved with palliative therapy; however, repeat MRI at 30 days in one of those three showed an interval increase in the extent of periventricular white matter reduced diffusion, although that patient's symptoms had improved (Figs. 5A, 5B, 5C, 5D, 5E, and 5F). In the four patients with carbon monoxide (CO) toxicity, follow-up MRI in two showed interval reduction in the extent of FLAIR hyperintensity (Figs. 6A, 6B, 6C, 6D, 6E, and 6F); notably, two of the three patients who improved clinically had an available presenting serum CO level of < 30%. The fourth had a serum CO of 45% and an ADC reduction of 62% and remained comatose. One patient had meningitis and renal failure after abusing various illicit drugs; any of a multitude of comorbid factors could have contributed to the toxic leukoencephalopathy, so the definitive cause was not identified because the patient improved. Finally, the cause of neurotoxicity in another patient was elusive and has not yet been ascertained because the patient was not on any known medications. The symptoms in that patient never entirely resolved; short intervals of clinical improvement after various vitamin therapies (vitamin B, folate, thiamine, and so forth) were followed by intermittent relapses.

Toxic leukoencephalopathy should be considered in the differential diagnosis of a patient who presents with recent onset of neurologic deficit and known exposure to a toxin that has been described as injuring the white matter [1]. A heterogeneous array of causes has been described that can lead to toxic leukoencephalopathy with similar clinical presentations [1-21]. In this article, we describe—to our knowledge, for the first time—a common appearance on DWI and on FLAIR imaging shared by these various entities. To improve recognition of the radiologic appearance and clinical presentation of this disorder, we have referred to this as “acute toxic leukoencephalopathy” on the basis of the accepted nomenclature that has referred to this clinical syndrome as “toxic leukoencephalopathy.” We have added “acute” to improve awareness of the DWI findings in the early phase of this potentially reversible neurologic syndrome.

Various medications can cause an acute presentation of toxic leukoencephalopathy. The clinical severity is variable; usually a mild reversible form occurs. However, the white matter alteration can lead to necrosis in the more severe cases, particularly when chemotherapy is combined with irradiation [2-7, 25, 26]. In cases secondary to chemotherapy, as illustrated in our three cases with histologic verification, the histologic findings may include demyelination, myelin pallor, myelin vacuolation, axonal spheroids, or macrophage infiltration, with necrosis considered the severe end point of the spectrum [25, 26]. Although no cases in this article were classified as a necrotizing leukoencephalopathy because no patients showed contrast enhancement on the initial examination, one patient did eventually show enhancement in the involved periventricular white matter later in the course of the illness and had evident necrosis on histologic examination [25, 26].

The basic pathophysiologic mechanisms leading to acute toxic leukoencephalopathy and reduced diffusion are unknown but are likely multifactorial. Chemotherapeutic agents such as methotrexate and 5-flourouracil may directly induce injury to the microvasculature but can also have indirect excitotoxic effects [27-30]. Immunosuppressive medications, such as tacrolimus and cyclosporine, may cause capillary endothelial injury, leading to blood-brain barrier dysfunction and occasional injury to the white matter [7, 31-33]. Metronidazole may induce free radical production from its derivatives [9, 10, 34-36]. Carbon monoxide and inhaled opiates may cause either toxic demyelination or spongiform degeneration of white matter that can progress with time to coagulative necrosis; the histology in reported cases of acute or subacute carbon monoxide and opiate toxicity can overlap with that of medication-related leukoencephalopathy [11-16, 37-41]. Hence, the cause of reduced diffusion in the periventricular white matter in acute toxic leukoencephalopathy likely varies with the cause of the disorder but may arise from intramyelinic edema and resultant myelin vacuolation, cytotoxicity via capillary endothelial injury, or direct toxic demyelination [5, 6, 28-30].

This study attempted to evaluate whether certain acute or subacute imaging markers (the initial extent of reduced diffusion, the initial extent of abnormality on FLAIR, or the percentage of ADC reduction) could predict the degree of reversibility of either the clinical neurologic deficits or the imaging abnormalities. However, we found that the extent of abnormality on the initial MRI did not correlate with the clinical outcome, whereas a moderate correlation did exist between the initial and residual or chronic extent on FLAIR. This observation underscores the importance of recognizing even mild or early cases of this disorder in order to prevent progression to a severe outcome. We think that the lack of positive correlation between the imaging and clinical outcomes, even with severe clinical symptoms initially, underscores the unpredictability of this syndrome and the importance of early recognition. However, we could not determine definite findings on MRI to predict reversible versus irreversible and severe outcomes in this study. Our review of the clinical records found that in some patients (particularly in each case of fludara bine toxicity), the symptoms presented nearly a month after the initial exposure; in others, MRI was not performed immediately after symptom onset, and patients continued to receive the causative medication. Hence, in this study, patients with an unknown cause or delayed presentation and subsequent continued exposure to the toxin might have lowered the sensitivity of ADC values to evaluate the severity. With greater clinical recognition of the finding of reduced diffusion in periventricular white matter in acute toxic leukoencephalopathy, the sensitivity of such thresholds could improve.

In four patients in this study, a cause of leukoencephalopathy was identified that had not been previously described to our knowledge; in all four cases, the white matter findings improved with correction of the offending cause. Two of these patients presented with acute hepatic failure and hyperammonemia and decreased diffusion in the thalami, a location noted to have elevated ¹³N-ammonia extraction on PET in cirrhotic patients; it has been suggested that the DWI abnormality may be related to swollen perivascular astroglial foot processes [42-46]. In the patient with hypertension who later developed cortical edema typical of posterior reversible encephalopathy syndrome and worsening periventricular white matter reduced diffusion, it is plausible that cytotoxic microvascular endothelial injury could have caused the DWI findings. Although posterior reversible encephalopathy syndrome typically favors the cortex or subcortical white matter, we posit that the reduced diffusion in the periventricular white matter due to hypertension could be similar to periventricular white matter abnormalities that have been described in patients receiving immunosuppressive (e.g. cyclosporine) therapy due to endothelial injury. Indeed, in two patients in this study cyclosporine was confirmed as the cause [7, 47]. In this regard, a recent study based on MR angiography and catheter angiography has shown that a vasoconstriction syndrome can occur in posterior reversible encephalopathy syndrome that results in reduced cerebral blood volumes in the affected areas; theoretically, similar hypoperfusion or endothelial injury could have resulted in ischemic demyelination in our hypertensive patient [48]. Because posterior reversible encephalopathy syndromes secondary to either primary hypertension or immunosuppressive therapy can appear identical to each other on MRI, it is plausible that, in a similar fashion, toxic leukoencephalopathy with reduced diffusion of periventricular white matter could have an appearance shared by both hypertension and immunosuppressive medications. This further raises a question as to whether an overlap exists between these two potentially reversible entities (posterior reversible encephalopathy syndrome and acute toxic leukoencephalopathy) that share similar causes [7, 47]. Regarding the patient with prolonged (> 1 month) hyperglycemia, the cause of reduced diffusion is unknown but could relate to microvascular or hyperosmolar injury [49, 50].

Exclusion of the three patients with HIE from analysis was based on this entity's not being typically classified as a toxic leukoencephalopathy. Also, this disorder generally results in permanent neurologic deficits if multifocal findings are present on DWI, although degrees of clinical and imaging reversibility have been reported in a few instances [15, 16, 19]. Our three patients with acute symptoms who were later found to have leukodystrophies were excluded from analysis because these are also not typically considered toxic causes of leukoencephalopathy but rather congenital genetic disorders; an acute presentation and a severe lesion extent herald a tendency not to normalize on follow-up MRI as well as a poor clinical prognosis [51, 52]. Regardless of whether these entities have some reversible component, this study showed that HIE and congenital causes can be discerned from the causes of toxic leukoencephalopathy by comparing the distribution of abnormalities on DWI with the distribution on FLAIR (Figs. 7A, 7B, 7C, 8A, 8B, and 8C). Also notable is that an ADC reduction of > 70% was seen in the affected white matter in each case of HIE, greater than was seen in any of the cases of acute toxic leukoencephalopathy in this study.

Our study has several limitations. First, although a large number of patients receive chemotherapy, only a small number are imaged with MRI; thus, it is difficult to know the true incidence of DWI abnormalities in the periventricular white matter in a retrospective study such as this. Second, there could be bias in the measured ADC values because areas listed as unaffected periventricular white matter could be affected microscopically but not visible on FLAIR or DWI. In this regard, ROIs were obtained from the most severely affected periventricular white matter on DWI and were compared with what appeared to be entirely unaffected periventricular white matter on FLAIR and DWI. Our observation of a significant difference between ADC values measured in the affected periventricular white matter versus the unaffected periventricular white matter on initial imaging (p < 0.0001), with no significant difference in those patients with follow-up imaging ≥ 21 days after presentation (p = 0.13), would suggest that it is valid to presume that the unaffected periventricular white matter may be relatively normal; however, only histologic evaluation could confirm this with certainty. Third, the grouping of unrelated causes of toxic leukoencephalopathy could be seen as problematic; therefore, there could be inherent difficulties in applying statistics to different causes of white matter abnormalities with potentially differing outcomes. However, this grouping has been accepted in the clinical literature, although it has not yet been recognized and grouped radiologically on the basis of the acute or subacute appearance on DWI and FLAIR MRI [1]. Fourth, a similar criticism could be raised regarding the MRI severity and clinical outcome because these scales are based on our review of the preliminary results of previous studies. We found such variation in degree of involvement, symptoms, and clinical outcome that we believed the best way to determine a correlation was to compare the relative degree of MRI and symptom severity at follow-up (if any). Fifth, a potential criticism is our use of different MRI field strengths because white matter lesions could be more conspicuous at 3 T and changes in field strength could affect ADC measurements. Some authors have noted no significant differences in measured ADC values comparing various regions of periventricular white matter, although others have noted small but statistically significant differences between regions and between observers [53-56]. This is one reason we used relative ADC measurements of affected white matter compared with unaffected periventricular white matter rather than using absolute ADC measurements. Future studies would optimally be prospective, use control patients, and be limited to a single scanner strength on the initial and repeat MRI examinations.

Additionally, one could potentially interpret our results differently regarding the lack of imaging correlation with the clinical outcomes. We have interpreted this to mean that imaging is critical early in the disease to prevent progression from continued exposure to the causative toxic agent. However, this lack of correlation may lead some to argue the opposite—that is, that imaging may not make a difference in the acute care of this disorder because the insult has already occurred and the disease will run its course. Although it is difficult to prove in a retrospective study such as this, we disagree with this interpretation because recent literature has suggested that worsening of both the periventricular white matter abnormalities (on T2-weighted and FLAIR imaging) and clinical symptoms is more likely to occur with increasing dose or repeat applications of the toxin, most typically described as resulting from chemotherapy [4, 57, 58]. It has been our preliminary observation that earlier intervention may lead to a dramatic reversal of initially severe MRI findings, and lack of recognition may lead to marked progression of initially mild MRI findings. We believe that these dramatic changes, encountered in several of our patients, most likely explain the lack of statistical correlation between imaging markers and clinical outcomes.

We believe that it is important to recognize early the syndrome of acute toxic leukoencephalopathy in patients presenting with acute symptoms for several reasons. First, it has been our observation that a minority of these patients continue to deteriorate and may die after continued exposure to a particular agent (Figs. 3A, 3B, and 3C illustrates one such patient), as described previously with chemotherapeutic agents such as methotrexate and with other nonchemotherapeutic toxins [1, 25]. Second, this syndrome may present with rather nonspecific but severe symptoms and can be confused clinically with other often irreversible syndromes (e.g., HIE); hence, the window for effective treatment may be missed. Third, in clinical scenarios in which these patients have a known underlying disease that can produce neurologic symptoms, such as metastatic disease, infection, organ failure, and so forth, prompt recognition of a toxic leukoencephalopathy can prevent inappropriate treatment and potential iatrogenic side effects. Hence, our opinion is that earlier imaging and recognition may improve outcomes in acute toxic leukoencephalopathy, but this can be proven definitively only by prospective long-term studies.

This preliminary study lists various toxic metabolic causes of acute leukoencephalopathy that can present with reduced diffusion in periventricular white matter and that can be reversible clinically and on MRI (with subacute HIE and leukodystrophies as mimics). However, there are a number of other nontoxic causes that were not encountered in this study but that could also theoretically cause reduced diffusion in the periventricular white matter and should be considered in the differential diagnosis. These include venous is chemia, neoplasm, prolonged seizures, periventricular white matter infarcts, infections, mitochondrial diseases, and other congenital metabolic or demyelinating diseases [20, 21, 59]. Comparing the distribution of the abnormality on DWI with the distribution on FLAIR images can potentially narrow the differential; for example, the accompanying cortical hyperintensities on FLAIR distinguish HIE from the solely periventricular white matter findings of acute toxic leukoencephalopathy. We note that this potentially reversible entity should not be confused with posterior reversible encephalopathy syndrome, another reversible syndrome that can arise from a variety of causes similar to those listed here (e.g., chemotherapy, immunosuppressives, and hypertension) [22, 23]. Several factors enable distinction of this entity from posterior reversible encephalopathy syndrome—most notably, posterior reversible encephalopathy syndrome typically affects the cortex or subcortical white matter on FLAIR; the periventricular white matter immediately around the ventricle is not involved except in severe cases in which the subcortical white matter is already involved. In addition, posterior reversible encephalopathy syndrome only uncommonly involves reduced diffusion [22, 23].

In conclusion, this study describes various potentially reversible causes of acute toxic leukoencephalopathy that can present with symmetric reduced diffusion in the periventricular white matter and can be discerned from subacute HIE or congenital genetic disorders by comparing the initial DWI with FLAIR images. These preliminary results suggest that the DWI findings can be entirely reversible and that the extent of abnormality on the initial FLAIR images can modestly predict the chronic extent on FLAIR but cannot reliably predict the clinical outcome. This underscores the need for prompt radiologic identification of this syndrome to alert the ordering physicians to potential causes because the MRI findings and clinical symptoms can entirely resolve with early recognition and appropriate therapy.

Address correspondence to A. M. McKinney ([email protected]).

CME This article is available for CME credit. See www.arrs.org for more information.

This article is available for CME credit. See www.arrs.org for more information.