Previously presented in part at ECVN congress, September 2007, Bern, Switzerland.
Corresponding author: Peter M. Smith, Department of Veterinary Science, Small Animal Teaching Hospital, University of Liverpool, Leahurst, Chester High Road, Neston, CH64 7TE, UK; e-mail: firstname.lastname@example.org.
Background: The optimal treatment for meningoencephalomyelitis of unknown etiology (MUE) remains unknown, despite the widespread use of a variety of immunosuppressive drugs.
Objective/Hypothesis: To compare the efficacy of prednisolone combined with either vincristine and cyclophosphamide (COP group; n= 10) or with cytosine arabinoside (AraC group; n= 9).
Animals: Nineteen dogs with neurological deficits, neuroimaging, and cerebrospinal fluid abnormalities consistent with a diagnosis of MUE.
Methods: Prospective, blinded, and randomized clinical trial. Dogs fulfilling the inclusion criteria were randomly allocated to receive 1 drug regimen.
Results: Four of 10 dogs in the COP group and 5/9 in the AraC group survived > 12 months but neither the survival time nor the time-to-treatment failure differed between the 2 groups. Treatment with COP resulted in an unacceptable incidence of adverse effects.
Conclusions: The adverse effects of COP make it an unsuitable treatment for MUE. Although survival of animals treated with AraC was broadly similar to that reported in recently published studies describing this treatment, it remains unclear whether it confers any benefit over using prednisolone alone.
Meningoencephalomyelitis of unknown etiology (MUE) is the term recently coined for clinical cases in which magnetic resonance imaging (MRI) and cerebrospinal fluid (CSF) changes indicate inflammatory central nervous system (CNS) disease but for which there is no histopathological confirmation of the diagnosis. Differential diagnoses for MUE include granulomatous meningoencephalitis (GME), necrotizing meningoencephalitis (NME), necrotizing leukoencephalitis, idiopathic tremor syndrome, infectious meningoencephalitis, and neoplasia. After testing for infectious and neoplastic diseases, most cases of MUE are assumed to result from dysregulation of the immune system and are typically treated with immunosuppression. Several treatments have been used for MUE,1–8,a yet the different treatments have not been directly compared with each other in randomized trials. This lack of clinical trial data means that recommendations regarding treatment for this condition are weak, consisting of little more than anecdote.
There are several obstacles to constructing large scale clinical trials for MUE, most notably the variable nature of the condition, meaning that robust entry criteria and outcome measures can be difficult to define. Secondly, MUE often results in affected animals being euthanized but, in common with many other diseases of domestic animals, its timing will vary among attending veterinarians, which can impact the utility of survival as an outcome measure. There is therefore a pressing need to develop alternative means of outcome assessment.
This current study was devised to address the lack of comparative data by constructing a comparison between 2 therapies for MUE in a “head-to-head” randomized trial. It was conceived as a phase II trial in which a small number of typical cases would be treated randomly and the data then used to establish guidelines for a future large-scale phase III trial. Therefore, in this study we aimed to achieve several objectives: (a) to estimate the magnitude of the difference between 2 different treatments to derive power calculations for future large scale studies; (b) to identify obstacles to carrying out large scale clinical treatment trials on this condition; (c) to test a neurological scoring scheme that could be used to avoid having death or euthanasia as the outcome measure; and (d) to test against a widely used protocol a novel therapeutic protocol that had been anecdotally successful in our hands.
We elected to compare prednisolone combined with either cyclophosphamide and vincristine (COP protocol) or cytosine arabinoside (AraC). Drugs that suppress the immune system, particularly corticosteroids, form the mainstay of therapy for MUE. However, survival times in dogs treated with corticosteroids alone are relatively poor and relapses are common,2,7 prompting clinicians to adopt a variety of adjunctive therapies.1,3–8 Supplementing prednisolone with AraC, an antimetabolite used predominantly in the treatment of hemopoietic neoplasia, appears to show promise in treating dogs with MUE, with a number of reports documenting survival times that apparently outstrip those reported in previous studies using only prednisolone.3,6,8 Another regimen that anecdotally appears effective is the combination of COP alongside prednisolone. The rationale for this regimen is that GME shows histological features similar to lymphoma in some animals9,10 and because cyclophosphamide is widely used in treatment of other diseases with a suspected autoimmune etiology.11
Materials and Methods
All dogs were presented to the Queen's Veterinary School Hospital, University of Cambridge, between March 2004 and November 2006 for investigation of neurological dysfunction. All dogs underwent a thorough general examination and a neurological evaluation performed either by a board-certified neurologist or by a neurology resident. Dogs were considered suitable for inclusion in the trial if they had (1) evidence of focal or multifocal neurological dysfunction; (2) either an MRI scan or a myelogram consistent with inflammatory CNS disease; and (3) abnormal CSF suggestive of granulomatous inflammation (here defined as pleocytosis comprising at least 50% mononuclear cells). In all but 2 cases, the possibility of infectious diseases capable of causing granulomatous inflammation was specifically excluded by serological testing for evidence of Toxoplasma and Neospora infection; in the remaining cases, clinical signs were deemed too severe to postpone treatment pending further laboratory tests and treatment was therefore instigated without testing. Canine distemper virus testing was negative in 2 of only 3 dogs in which the CSF cell count was <75 cells/μL. Dogs were tested for other infectious diseases if any features of the case heightened the suspicion of an infectious process, such as exposure to ticks prompted testing for Anaplasma phagocytophilum and Borrelia; all tests were negative.
The study was intended as a phase II clinical trial to compare the efficacy of the 2 treatment regimens—both of whose safety has been previously demonstrated in animals—in a small group of patients.12 Two groups of 10 animals were included: treatment plans were placed in 20 sealed envelopes and cases were randomly allocated an envelope at the outset of treatment; 1 case in group 2 was retrospectively excluded from the study for failing to meet the inclusion criteria (the CSF sample contained around 90% neutrophils). There were 2 clinicians associated with each case, 1 to coordinate drug administration and deal with client and referring veterinarian communication and another to examine animals during the treatment and follow-up periods. The assessing clinician was blinded to treatment category by restricting access to any hospital records detailing specific treatments. This trial was approved by the Departmental Ethical Review Board.
Animals randomly allocated to group 1 (COP protocol) included 4 West Highland White Terriers, 2 Lhasa Apsos, 2 crossbreeds, 1 Airedale Terrier, and 1 English Pointer; 5 were male and 5 were female. Those in group 2 (AraC protocol) included 2 West Highland White Terriers, 1 Cairn Terrier, 1 Airedale Terrier, 1 Bassett Hound, 2 Labrador Retrievers, 1 Maltese Terrier, 1 Bichon Frise, and 1 Crossbreed; of these, 7 were male and 3 were female. There was no significant difference in age or weight of dogs included in the 2 groups (Table 1).
Table 1. Summary of clinical and clinical pathology findings.
MRI was performed under general anesthesia with an Esaote Vet-MR scannerb (0.2 T permanent magnet); typically this included T1- and T2-weighted images in the sagittal and transverse planes, a FLAIR image in the transverse plane, and T1-weighted images after administration of gadobenate dimegluminec at 0.1 mmol/kg. Myelography was performed with iohexold administered via cisternal or lumbar puncture. Cerebrospinal fluid was obtained by either cisternal or lumbar puncture and during the same anesthetic episode as neuroimaging. CSF samples were analyzed on site on the same day as sampling, including a total cell count, morphological examination, and measurement of total protein concentration.
Animals were randomly allocated to receive the following treatments:
Group 1—low dose lymphoma “COP” treatment protocol, with vincristinee (0.5 mg/m2 IV, every 7 days for 8 weeks, then every 14 days), cyclophosphamidef (50 mg/m2 per os, every 48 hours for 8 weeks, then the same regimen given in alternate weeks) and prednisoloneg (40 mg/m2 per os, every 24 hours for 7 days, then 20 mg/m2 every 48 hours for 7 weeks, then the same regimen given in alternate weeks). In some cases, because of the size of dogs and the formulation of the tablets, cyclophosphamide was administered less frequently than every 48 hours but the total dose received each week remained the same.
Group 2—prednisolone was administered according to the same protocol as in group 1, along with cytosine arabinosideh (AraC) instead of cyclophosphamide and vincristine. AraC was given once only at the instigation of therapy, by IV infusion over 24 hours, diluted in Hartman's solution; a total dose of 100 mg/m2 was administered. No further treatment with AraC was performed.
For dogs in both groups, the dose of prednisolone was tapered to suit individual requirements after 6 months and stopped if possible.
Assessment of Clinical Signs
A scoring scheme was designed (Appendix 1) and retrospectively applied to the results of the neurological examination. The scoring scheme was applied by an individual (DS) blinded to the nature of the trial and to the groupings of the animals; this outcome analysis was therefore double blinded.
The mean and median age and weight of the animals, the CSF cell count and the CSF protein concentration were all calculated by standard statistical software.i After checking the distribution of data for normality by the D'Agostino-Pearson Omnibus K2 normality test, the age, and weight of animals in the 2 groups were compared by an unpaired t-test, taking P < .05 as the level of significant difference; CSF cell count, CSF protein concentration, and neurological scores at the outset of treatment in the 2 groups were not normally distributed and were compared by a Mann-Whitney U-Test. The survival time and time-to-treatment failure were documented by Kaplan-Meier plots, generated by SPSSj and the 2 groups were compared by log rank analysis. The proportion of animals surviving at 1 month and 12 months was also calculated; this was compared by Fisher's exact test to calculate a two-tailed P value.
MRI scans typically showed 1 or more lesions that were hyperintense on T2-weighted or FLAIR images. In 1 dog, there was no parenchymal lesion to explain the neurological deficits but there was hyperintensity of the meninges on T2-weighted images and marked enhancement on T1-weighted images after gadolinium administration; in 1 other dog that underwent MR imaging, no lesion was detected. In animals with identifiable parenchymal lesions, these were seen in the brainstem/cerebellum only in 1 dog, in the cerebrum/thalamus only in 6 dogs, and at both locations in 6 dogs.
Of the 4 animals with spinal cord disease, 3 had lesions localized to the lumbosacral intumescence. In 2 of these dogs, myelography was performed and neither had evidence of spinal cord compression, but in 1 the spinal cord appeared swollen in the region where a lesion was predicted by neurological examination.
Cell counts and CSF protein concentrations were not significantly different between the 2 groups (Table 1). In 1 dog, the CSF contained a predominance of morphologically uniform cells, suggesting the possibility of lymphoma; however, subsequent immunostaining and analysis by flow cytometry showed a nonclonal population of lymphocytes.
For survival analysis, all 7 dogs alive at the time of writing—4 in group 1 and 3 in group 2—were censored. Five further dogs were censored: three from group 2, including 2 that were euthanized for non-neurological disease (lymphoma and pneumonia) and 1 whose treatment was changed after a relapse but which nevertheless survived over a year receiving only prednisolone, and 2 from group 1—both of whose treatment was stopped, 1 for financial reasons and 1 because of myelosuppression (Fig 1). Median survival was estimated at 1,063 days in group 2 but could not be calculated in group 1 because of the large number of censored cases. Using an intention-to-treat analysis showed a median survival time of 198 days in group 1 (95% confidence interval [CI], 247–914 days) and 735 days in group 2 (95% CI, 195–1,274 days; P= .89).
The proportion of animals surviving 1 and 12 months was then calculated, excluding those same animals censored in the Kaplan-Meier analysis, to leave 8 animals in group 1 and 7 animals in group 2. In group 1, 5/8 dogs were alive at 1 month and 4/8 were alive at 12 months (50%) and in group 2, 5/7 were alive at both 1 and 12 months (71%). Thus, of the animals alive 1 month after initiation of treatment, 80% (4/5) in group 1 and 100% (5/5) in group 2 went on to survive 12 months or more. The number of animals alive at these 2 time points was not significantly different between the 2 groups (1 month, P= 1.00; 12 months, P= .61) (Fig 1).
Treatment Failure Analysis
To permit a more objective assessment of the response to treatment, a post hoc analysis of outcome was performed by scoring the neurological deficits (Appendix 1). Treatment was considered to have failed if the neurological score worsened, if there was a relapse after an initial improvement or if the animal died or was euthanized because of its neurological disease. The mean score at the outset of treatment was not significantly different between the 2 groups. Time-to-treatment failure differed from survival time for 3 animals in group 1 (1 which deteriorated at day 27 but survived until day 40, 1 that deteriorated at day 6 but survived until day 15 and another which suffered a relapse at day 632 but remained alive for more than 1,051 days) and 1 animal in group 2 (which worsened at day 40 but ultimately survived for 376 days). A Kaplan-Meier plot of time-to-treatment failure (Fig 2) revealed a median value of 632 days in group 1 (95% CI 0–1,342 days) and 1,063 days in group 2 (102–2,023); log rank analysis showed no significant difference between the 2 groups (P= .74).
All dogs showed varying severity of corticosteroid-related adverse effects, including polyphagia, polydipsia, and polyuria. A number of additional adverse events considered to be related to treatment were also identified. In group 1, 1 dog became myelosuppressed after its 1st treatment and 2 developed hemorrhagic cystitis, 1 after 3 months and the other after 5 months. One further dog developed pyometra after 4 weeks of treatment; this was successfully treated but cyclophosphamide was subsequently discontinued because of periodic lymphopenia and mild gastrointestinal disorder associated with this drug. In group 2, no adverse effects were reported that could be definitively linked to the treatment regimen. However, 1 dog was euthanized after developing Bordetella sp. bronchopneumonia and another was euthanized after being diagnosed with lymphoma. Additionally, 1 dog had an acute deterioration in its clinical condition after initiation of treatment, though this resolved over the following 48 hours.
MUE comprises a number of different diseases, most of which can be excluded on the basis of laboratory testing for infectious diseases and neoplasia. The largest subset of MUE is likely to be GME, a relatively common neurological condition for which the etiology remains obscure but in which histological changes and immunohistochemistry together suggest an immune-mediated mechanism.13–15 Treatments for GME and MUE have therefore focussed on immunosuppression, predominantly with prednisolone but, more recently, supplemental immunosuppressive drugs have also frequently been employed. The range of drugs that have been tested is extensive and continues to grow.1–8,a However, none has been adequately assessed in comparative clinical trials and therefore no 1 treatment can currently be unequivocally recommended. In this study, we compared prednisolone with either AraC or COP as an adjunct and found no significant survival benefit in using 1 protocol over the other. However, the incidence of adverse effects in the animals treated with COP was unsatisfactory, indicating that this treatment should not be tested further in future clinical trials.
The mean and median survival times could not be compared for the 2 groups in this study, because several animals in each group were still alive at the time of writing. Another outcome measure, survival 12 months after diagnosis, was found to be 50% (4/8) for COP-treated animals and 71% (5/7) for AraC-treated animals. This corresponds relatively well with recently published retrospective studies examining the survival of animals treated for MUE by various treatment regimens4–8 and indicates that the prognosis for dogs with MUE is perhaps better than earlier studies on GME suggested.2 Interestingly, only 1 animal that survived to 1 month failed to survive 12 months. Furthermore, those animals that survived upto a year often lived for a relatively long period beyond this—>3 years in 2 dogs—suggesting that animals alive at 1 month may have a relatively good chance of living several more years.
Choice of Medication
The COP protocol is a combination that has not been reported previously for treating MUE (or GME) but is a regimen that anecdotally appeared beneficial. It is widely used in the treatment of animals with lymphoma and there is a substantial body of evidence demonstrating its safety and it is therefore well placed for testing in a phase II therapeutic trial. Ideally, any new treatment would be tested against a placebo or against the previous “gold standard” treatment, in this case prednisolone. However, this approach was difficult to justify in the face of widespread opinion and recent reports in the literature that prednisolone alone is not a satisfactory treatment for MUE.1,2,7 For this reason, AraC was adopted as comparator, though it was administered only once so that the regimen was as close as possible to prednisolone alone.
The route of administration of AraC in the current trial was different from that described previously in that we used an IV infusion over 24 hours instead of the widely reported 4 SC injections, 12 hours apart over a period of 48 hours.3,6,8 This was chosen because AraC is rapidly metabolized to inactive uracial arabinoside16 and the elimination half-life after IV injection in dogs is only 69 minutes.17 Studies in human subjects indicate similar pharmacokinetics after SC injection18 and, because AraC needs to be present for a protracted period to be effective,19 this might not be an effective route for maintaining therapeutic concentrations. Even our regimen of 100 mg/m2 infused over 24 hours may not have been sufficient to achieve and maintain the minimum in vitro cytotoxic concentration of 1 μM within the CNS.20 A steady-state CSF concentration of 8.3 μM was only achieved in experimental animals by using a dosage rate 12 times that used in the current study.17 Furthermore, the CSF concentration is likely to be different from that in the brain and spinal cord: experimental work has shown that tissue levels of AraC are high only at the interface between CSF and brain but decline exponentially into the parenchyma.21 Given the angiocentric nature of the disease,13,14 it might still be possible to achieve a therapeutic effect at the required site of action in most animals, although the complex pharmacokinetics of this drug mandate caution in choosing the optimal dose and means of administration.
A number of adverse effects were identified in both groups of animals. In the COP group, one 4-year-old dog developed pyometra 4 weeks after the inception of treatment, which is relatively young for this condition.22,23 Two dogs in this group had myelosuppression, one of which also had signs of gastrointestinal disease association with cyclophosphamide. Another major concern with animals in the COP group was the development of hemorrhagic cystitis, which occurred in 2 of 4 animals surviving to 6 months. This is a recognized complication of treatment with cyclophosphamide, arising predominantly through the irritant effect of a metabolite, acrolein, on the bladder epithelium24 and is both debilitating for the animal and difficult to treat. The incidence of hemorrhagic cystitis in this group of animals was higher than has been reported in dogs given COP for treatment of lymphoma25 and in humans26 and might simply reflect an unhappy coincidence. However, in the face of concerns about myelosuppression and a failure to demonstrate improved survival in COP-treated animals, this drug combination cannot be considered a prudent treatment for MUE.
In the AraC-treated animals, few adverse effects other than those expected through the administration of high doses of corticosteroids were encountered. Previous studies have revealed that relatively high doses of this drug (600 mg/m2) are well tolerated,17 though myelosuppression and mild gastrointestinal upsets have been reported at this dose level,27 and previous reports describing SC administration of AraC to treat MUE have reported few adverse effects.3,6,8 In the current study, 1 dog developed Bordetella pneumonia after being exposed to other dogs with kennel cough, which is an unusual sequel to upper respiratory tract infection and seems likely to be related to immunosuppression. In another, a diagnosis of lymphoma was made around 2 years after initiation of treatment. A higher incidence of lymphoma is well recognized in humans given immunosuppressive treatments28–30 and seems likely to be a risk in dogs treated similarly. However, lymphoma is a common neoplasm in dogs and a link to AraC seems unlikely on the grounds that only 1 dose of the drug was given. Nonetheless, its occurrence—and that of pneumonia and pyometra—serves as a reminder that cases should be managed vigilantly.
Older studies of MUE (or GME) tend to report shorter survival times than do more recent studies, in which multidrug treatment regimens tend to be used. For example, an early report of 42 dogs with histologically confirmed GME showed a median survival of 14 days,2 whereas recent studies of animals with MUE treated with prednisolone and either procarbazine or AraC describe median survival in excess of 12 months.6,7 This could be interpreted as representing an improvement in the way in which MUE is now managed; however, for a number of reasons this view should be tempered. In the earlier study, all dogs had histological confirmation of GME and most of the cases had a postmortem diagnosis (35/42), thereby selecting for animals that had died (ie, those with a poorer outcome); despite this, one of the cases with focal GME survived 1,215 days, which compares favorably with current survival times.2 In addition, 12 dogs in this study died before treatment could be initiated, which dramatically skews median survival times downwards. To compare this study with a retrospective analysis in which animals do not have histological confirmation of their disease and in which all have been treated, is not therefore a reasonable comparison.
Only 1 animal in the whole study that survived to 1 month failed to survive 12 months. This indicates that animals surviving the initial phase of the disease have a good chance of a protracted survival and suggests that future studies should predominantly focus on improving disease remission. Because the most suitable drugs for inducing remission and preventing relapses might be different, trials to optimize each are probably best carried out independently, with an emphasis on outcome after a relatively short period of treatment likely to yield the most benefit to affected animals.
Measuring outcome in clinical trials is difficult. Long-term survival can be misleading, partly because it is influenced by both disease remission and maintenance and partly because it is subject to a variety of confounding influences, including client and pet character, veterinary advice, development of concurrent medical problems, and financial considerations. Subjective assessments of clinical improvement by both clinicians and owners, though valuable, are often inconsistent and unreliable and it is therefore incumbent upon researchers to establish a reliable scoring system that can be applied objectively. For this reason, we tested a system of assigning values to various neurological deficits in the current study and calculated an overall “neurodisability” score. Using this approach permitted us to calculate a time-to-treatment failure that eliminated some of the problems of relying on survival time, though we were still unable to identify a significant difference between the 2 treatment groups. However, this easy-to-apply scheme could readily be used in the future to compare both time-to-treatment failure and the magnitude of the response to treatment, which should allow novel treatments to be tested objectively over a relatively short time scale.
Two further problems that need to be addressed in future studies are the number of animals that need to be included and the establishment of a “gold standard” treatment against which new treatments can be tested. The current study included insufficient animals to determine a clear difference between the 2 treatment groups but on the basis of our results, a power calculation can be performed to estimate how many animals need to be enrolled in a study to detect a clinically significant difference, should one exist. Using this data, ∼85 dogs would have been required in each group to demonstrate a difference in survival at 1 year between COP and AraC/prednisolone treatment (50 and 71%, respectively, in the current study) with a standard power of 80% and P of .05. Given the adverse effects encountered with COP treatment, such a large-scale trial is clearly unnecessary; however, these figures also indicate that studies described in the literature are underpowered to determine a statistically significant difference in outcome. Another important problem is the inclusion of a placebo group. Previous studies indicate that animals with MUE (or GME) that receive no treatment fare dismally2,7 and to deny treatment is therefore not ethically acceptable. There is also a widespread view that prednisolone alone is not a satisfactory treatment for MUE or GME and this has prompted many to use a variety of immunosuppressive drugs alongside corticosteroids. In the current study, we obtained relatively good survival times with prednisolone combined with only a single treatment of AraC. While some studies have suggested a need for ongoing AraC treatment, our observations indicate that at least a proportion of animals with MUE can be treated with a moderate dose of corticosteroids alone and that future studies should examine this possibility.
a Sturges BK, LeCouteur RA, Gregory CR, et al. Leflunomide for treatment of inflammatory or malacic lesions in three dogs: A preliminary clinical study. J Vet Intern Med 1998;12:207.
b Esaote Vet-MR scanner, Esaote Vet-MR, Reading, UK
c Multihance, Bracco UK Ltd, High Wycombe, Bucks, UK
d Omnipaque 240 mg/mL, Nycomed Amersham Imaging, Amersham, Bucks, UK
e Vincristine Sulphate (1 mg/mL), Mayne Pharma Plc, Leamington Spa, UK
f Endoxana (50 mg tablet), ASTA MedicaLtd, Cambridge, UK
g Prednicare, Animal Care Ltd, York, UK
h Cytarabine (100 mg/mL) Mayne Pharma Plc
i GraphPad Prism 5.0 for Windows GraphPad Software, La Jolla, CA
j SPSS for Windows, version 15, SPSS Inc, Chicago, IL
The authors are grateful to Mike Herrtage for help in interpretation of MR images and Sophia McMillan for help in contacting owners.
Grant support: none.
Appendix 1. Summary of Neurodisability Score
The neurological scoring system was developed in consultation with a number of board-certified neurologists to provide an objective assessment of clinical status. The system allocates arbitrary scores to clinical deficits that can readily be identified in a routine neurological examination, the sum of these providing an overall disability score. Animals with more severe neurological deficits score more highly, hence plegia scores more highly than nonambulatory paresis, which in turn scores more highly than ambulatory paresis.