Clinical signs and follow-up
Knuckling was used as an inclusion criterion for definitive cases of AEP while obvious ataxia was an exclusion criterion, according to earlier observations in Sweden and Norway [1, 2]. In equines, the combination of pelvic limb knuckling and no obvious ataxia is rather unusual. In a report from Finland on this disease, one knuckling horse also had a moderate to severe ataxia potentially referable to a cervical spinal cord lesion . Such ataxia was not observed in any of the horses from our AEP-affected farms. Eight plausible cases in the present study were acutely and severely affected, some without registration of knuckling before recumbency. Because they appeared temporally close to more typical knuckling cases on the farms, they were likely AEP-affected; however, autolysis prohibited final diagnosis at necropsy. Hanche-Olsen et al. (2008) also reported paraplegia and recumbency within hours of onset in some cases .
The tests that most consistently produced typical pelvic limb gait deficits in cases were walking in circles, with and without tail pull (sway reaction test) and abrupt stop from trot. Caution is advocated regarding provoking movement in horses with severe AEP, as they may fall, as occurred in one case. On the other hand, mild cases were identified and 6 cases had no signs on the day of neurological examination, which demonstrates that systematic and repeated examinations are of value in an affected yard. In our experience, physical exercise on deep or uneven surfaces is sometimes necessary to elicit neurological deficits (knuckling) in mild cases or early stages of AEP. Our results indicate that the severity of the neurological signs observed is partly related to when in the disease process the horse is examined. Whether or not the variability also reflects a dose-related toxic aetiology remains to be investigated.
The lesion found in AEP has been described as a demyelinating polyneuropathy [1, 3] and neurological signs other than knuckling are theoretically possible but other signs were uncommon and inconsistent. Hanche-Olsen et al.  described decreased sensory perception of skin distal to the hock in a few cases but this was not typical for AEP in the present study. The decreased response to skin pricking in the front limbs in 2 cases might have been due to the stoic nature of these individuals. Repeated examination at another time would have been valuable. Seven case horses and 2 controls had mild focal atrophy of muscles, which is a possible feature of neuropathies. The mild deficit of muscle mass in the croup and back in these cases could also well be an effect of lack of activity during their convalescence. There were no signs of atrophy of digital extensor muscles in the pelvic limbs, which would be the expected location of a possible neurogenic atrophy in a motor neuropathy causing knuckling. Since the 2 controls were not showing any other signs attributable to neuromuscular disease, the mild muscle loss noted was not interpreted as an effect of polyneuropathy. Muscle biopsies may have elucidated this further.
The 29% mortality rate in horses with AEP and loss of horses in 8 of 13 farms in this study demonstrate that the condition often becomes severe and life-threatening. Including plausible cases, mortality was greater in horses older than 5.5 years. The most severely affected pony that survived in the present study was salvaged by repeated lifting with slings. This can be contrasted to a previously reported grave case, similarly supported for 8 weeks and then subjected to euthanasia due to recumbency . In severe and rapidly progressing cases, the prognosis must be guarded, whereas nonrecumbent cases have a good long-term prognosis. The times to cessation of knuckling (median 4.4 months) and to return to work (median 6.6 months) can only be considered an approximate measure of return to health because of the variable ability and keenness of owners for observations of knuckling.
Epidemiology and risk factors
The observed uneven geographical dispersion of AEP remains enigmatic. Only Nordic countries have reported AEP and Norway has most cases both in absolute and relative numbers. During 2007–2009, several-fold higher number of farms and horses were reported with AEP in Norway (10 included farms and 25 other farms) compared with Sweden (3 included farms and 4 other farms), despite the fact that the Swedish horse population is estimated to be 6 times larger than the Norwegian population (Sweden, n = 362,700 in 2010 ; Norway, n = 60,000. K. Hustad, personal communication).
The demography as well as the broad spectrum of breeds and usages affected with AEP in the literature infer that AEP is not caused by a genetic trait nor linked to gender and not related to a specific usage of horses. The clinical examination did not detect signs of general infection nor any localised signs from other than the locomotor system, specifically, signs of brain or cranial nerve dysfunction were not present. The clinical and epidemiological findings may fit with a toxic or toxicoinfectious aetiology of AEP involving exposure of several horses on a farm to a common factor, possibly in the forage [1-3]. Immune-mediated reactions may be involved in the pathogenesis in conjunction with a microbial or chemically derived toxin. Further studies on the pathogenesis are needed.
Coincidence of feeding wrapped forage and AEP was demonstrated in the present study, in accordance with most earlier reported cases in Norway and Sweden [1, 2]. The first outbreaks of the knuckling syndrome appeared in Norway and Sweden in the 1990s, concurrent with a shift from hay to wrapped forage on many horse farms. Wrapped forage is now used in over 50% of all horse farms in Sweden and the features of this type of forage have been previously described . An interesting finding in the present study was that cases appeared only among horses fed certain batches of forage on 2 of the studied farms. Temporal and spatial coincidence does not prove a causal relationship and despite the feeding of wrapped forage to all horses in the affected farms in the present study, all were not affected with signs. The earliest cases on the studied premises were often the most severe in this case series. One may speculate on the possibility that there may have been only a short exposure to the putative toxin in some farms, with the horses receiving the highest doses appearing affected first and most severely and the horses with lower exposures taking longer to show up and often presenting milder signs. However, putative toxic factors in forage may be unevenly distributed and sensitivity may differ among individuals, which complicates the interpretation. In contrast to our findings, there are also a few historical observations of AEP cases occurring in late summer–autumn [1, 2], also in Finland with different feeding management . Thus, the hypothesis of a toxic factor in forage remains unproven at this stage. A potential coupling between food poisoning and polyneuropathy exists in man, where the immune-mediated Guillain–Barré syndrome has been associated with exposure to pathogens such as Campylobacter jejuni . The authors are currently carrying out further evaluation of features and management of wrapped forage as risk factors for AEP, as well as neuropathological studies to elucidate the role of the immune system in AEP.
Horses 12 years or older were less associated with AEP than were younger horses. On the other hand, young cases (<5.5 years) had a higher survival rate than older horses and a tendency to return to work in a shorter time. Box stabling at night, compared with loose-housing tended to be less associated with AEP in the multivariable analysis. The interpretation of these findings is unclear, but may be facilitated by increased knowledge on aetiology and pathogenesis. Weight loss before the farm visit was associated with AEP cases. In retrospect, as bodyweight was subjectively estimated by the owner without weighing the horses regularly, we suggest that little emphasis should be put on this finding. Low work intensity was associated with AEP cases in the univariable analysis but due to many missing observations, it was not formally tested in the final multivariable analysis. Thus, correlation to other variables cannot be analysed.
The farms included in the present study represented around 30% of the farms that reported having AEP during the period. Inclusion of more farms would have strengthened the study but the sampled farms (type, size, time of year) seem to be representative of the typical affected population. The power in the case–control part was low and inherent was that nonaffected horses were part of affected herds. Compared with the power calculation if a reasonable design effect of 2 would prevail, then a sample size twice the current size would be needed to find significant differences.
Considering the prospective approach used in the present study, there was a relatively large amount of missing data; however, data on missing variables were actively sought and updated to a large degree. In the multivariable analyses there were varying numbers of observations for each variable and in the final model 148 of 157 horses were included. The approach made it possible to use most of the data in the final conclusions. This means that all models in the analysis were not done on the same dataset, i.e. during model reduction the models were not truly hierarchical. Reduced datasets did not alter the conclusions (data not shown).