In the present study we used an attentional visual-spatial task with central cues, the so-called central cue Posner's paradigm, in order to investigate the altered cognitive processing in two subtypes of MS patients and controls. At the same time we explored the nature and pattern of these cognitive abnormalities and their relationship with clinical variables using a multivariate statistical approach.
MS patients demonstrated a slower response speed than controls when detecting the mere presence of a stimulus previously cued (attentional orienting) as well as when they had to detect a stimulus which appeared over the opposite cued site (attentional reorienting).
Our results replicate other studies in healthy subjects, which have observed an increase of RT for an invalid cueing condition compared with a valid cueing condition [25, 26, 34]. Orienting to cues provides information about where the target will occur, with consequent benefits for RT. It is expected that the processing of a target stimulus that is not previously cued, takes longer, or causes more errors, than that of a cued target stimulus. Control and RRMS groups showed this "valid effect", but patients of the BMS group did not.
Results obtained by BMS patients cannot be easily attributed to a deficit in the disengaging of attention, since these patients obtained a similar RT in valid and invalid condition. Their performance could be attributed rather to a deficit in engaging attention or follow-up of cue, since they seem to deal with all stimuli equally. Similarly, these differences cannot be explained by a speed-accuracy trade-off because MS groups were also less accurate than the control group. Both RRMS and BMS groups had significantly lower CR percentage and more errors than the control group, with predominance of the missed type. In patients with BMS, the missed type error makes up the vast majority of lack of accuracy during the task. Since we have considered an error type missed when no response to the target stimuli occurred, or if the response occurred 700 ms after target stimuli, then the lack of accuracy of patients with BMS could be interpreted as a slowed cognitive processing on attentional tasks.
Slower reaction times in MS patients have been frequently reported [35, 36] as well as frequent attentional deficits [37, 38], and a slowing of mental processing independent of motor slowing has been suggested [37, 39, 40]. This slowing of the speed of mental processing and other cognitive dysfunctions can already appear early in the natural history of the disease, or even in clinically isolated syndromes suggestive of MS [41, 42]. It can also be predictive of global cognitive decline  and is particularly pronounced with visual and auditory tasks [37, 40, 43].
The findings of our study suggest that subtypes of MS disease may be associated with a moderate/severe attentional impairment and with information processing deficits. Compared to RRMS patients, BMS patients were slower and less accurate, suggesting that BMS patients could incur greater information processing impairments. These deficits could be affecting attentional functions in terms of attentional orienting and reorienting affecting the BMS disease subtype more than other subtypes of MS (such as RRMS) despite an apparent milder physical disability.
Although our results have clearly shown that BMS patients were slower and had less accuracy in visually presented information processing than RRMS patients, there are very few studies (most of them very recent) that have explored or found evidence of altered cognitive processing in the benign subtype of MS [9, 10, 44]. Furthermore, very few functional studies have tried to assess different aspects of attentional processes and information processing in MS patients with a benign profile, using cognitive paradigms like the Posner paradigm  or the Stroop paradigm .
Both RRMS and BMS patients may use different cognitive strategies in task performance, although this suggestion was not checked. For instance, some studies have demonstrated that when patients with MS are provided with additional time to process information, they perform as accurately as controls .
Event Related Potentials
The analyses of the ERP data also revealed some interesting findings. In combination with these behavioural data, electrophysiological results allow us to propose the existence of distinct levels of altered stimulus processing in MS patients, notably in attentional mechanisms and different stages of information processing.
CNV period and ERP Amplitude
The topographical distribution of eCNV clearly separates BMS patients from controls, but not from RRMS patients. When regions are considered separately, eCNV is smaller (amplitudes more positive-going) in BMS patients than in controls at central and parietal but not frontal sites. On the other hand, a striking finding of the present study has been the more negative amplitude of cCNV period at fronto-central sites that RRMS patients showed in comparison to the control group but not the BMS patients.
According to the functional hypothesis of early period of CNV [15, 46, 47] current results may indicate a reduced or worse activation of orientation and preparation mechanisms in BMS patients, after the presence of a cue. RRMS patients, however, might be exhibiting an increased attention at the beginning of the preparation stage or greater task motivation since the amplitude of CNV increases (more negative) with the force required to make a response, as well as when more attention is directed to the response [48, 49]. These amplitude abnormalities found in different temporal periods of CNV both for BMS and RRMS patients may therefore be revealing diverse neural mechanisms involved during preparation performance, which may be affected in different degrees. However, both MS groups appear to have the amplitude of the final period of CNV intact, which is traditionally related to sensory preparation for the imperative stimulus as well as to preparation for the motor act.
Consistent with a large body of evidence, the CNV is a sustained slow potential that develops during the interval between two task-relevant stimuli, with the second stimulus usually requiring a motor response [50, 51]. After the initial response to the warning stimulus is completed, an early negativity or eCNV develops over prefrontal, precentral or parietal areas depending on the task or paradigm used [52, 53]. Specifically, a change on eCNV amplitude has been observed when a spatial cue is presented in comparison to when the cue is not presented. This finding has been functionally interpreted in terms of activation of an executive mechanism controlling orientation or attention to a stimulus, assigning specific neural resources to the cued side and the posterior tCNV and suitable preparation responses [46, 52, 53]. In general, the main contributor to the CNV is the activation of fronto-parietal networks indicating the endogenous attentional effort during the CNV period, and the activation of the task-related neural set [15, 16, 47, 54].
Regarding BMS patients, our findings related to CNV seem to demonstrate that BMS patients, with the slow advance of disease, could in turn develop deficits in the fronto-parietal networks involved in tasks with cued stimuli. When adopting the view that early CNV activity (starting around 500 ms after cue onset) is indicative of the start of cue orientation processes, whereas the later CNV wave reflects expectation of the response and target location, it could be speculated that BMS patients invest less effort or have a compromised attentional mechanism in cue-orientation processes. Some of the previously discussed cortical alterations found in MS patients, which are linked to neural generators of CNV, have been partly reported in recent studies with BMS patients [9, 10, 45] and could be the reason that BMS patients showed cognitive abnormalities beyond the behavioural slowing that we have highlighted.
RRMS patients, however, appear to have good preservation of the brain mechanisms indexed by the CNV, showing even greater amplitude during the specific time period of CNV. Previous studies have shown that the CNV amplitude may be more negative when more attention is focused on the task [46, 48] and it is jointly determined by the two processes of attention and arousal . Therefore, RRMS patients could be mobilizing a greater amount of neural resources during initial periods of CNV (around mean 800 ms after cue onset) related to expectation and preparation for cognitive and behavioural responses.
Furthermore, this finding may also be connected to other striking results related to the higher amplitude of the LN component that RRMS patients obtained in comparison to BMS patients. However, any interpretation regarding this finding should be cautious since the only differences found were between the patient groups but not between patients and controls. Neuroimaging and electrophysiology studies have clarified some aspects related to the usual neural sources for the LN wave and other late ERP negativities during the performance of different conflict task, including spatial conflict [55–57], further indicating that the size of conflict is important in the generation of associated cortical activity.
A first interpretation for the largest amplitude of LN and cCNV in RRMS group could refer to an increase in cortical activity and neural resources associated to higher-order processing of stimuli. In this respect, the amplitude of long-latency of ERP components is considered as an index of brain activity that is required during a task, or proportional to the amount of attentional resources devoted to a given task [58–60]. Consequently, the higher LN amplitude for RRMS (and the lower amplitude for BMS) as well as higher cCNV, could also suggest the existence of compensatory cortical activation and brain reorganization when these patients perform attentional or memory tasks, in accordance with some functional neuroimaging findings during performance of the spatial tasks [61–64]. RRMS patients, therefore, could be manifesting a particular brain organization or increasing amounts of neural sources (indexed to higher ERP amplitudes) involved in specific time periods which relate to preparation performance and spatial conflict monitoring in the Posner paradigm.
However, some limitations of our study do not allow us to conclude this point of view. As far as we know, there are no other studies that have previously assessed the functional mean of CNV in MS, so we cannot compare our results in MS patients with other studies. Nevertheless, further studies should aim to explore and verify these aspects.
Regarding amplitudes of other ERP components analyzed in our study, we tried to find the electrophysiological correlates of the behavioural advantage of validly cued targets. Typically, the P1 amplitude is enhanced to valid cues. However, the control and the RRMS groups showed greater amplitude to the valid condition compared with BMS patients, suggesting that BMS patients seem to have a lack of validity effect to amplitude P1 . Some improvements, such as increasing the size of our sample, might be made in future studies to determine better electrophysiological correlates of behavioural valid effect.
In contrast to our current results, some authors have reported frequent decrease of ERP component amplitude in MS patients, but in particular for N2 and P3 waves and for the auditory modality. A decrease of amplitude for P3 [65, 66], for P2 and for P3 , for N2 and for P3 , for mismatch negativity complex and P3 , or for P2 and N2 waves [19, 69, 70] has been reported in MS patients. These results have been interpreted in terms of information processing difficulties and an impairment on attentional orienting mechanisms related with a disruption of the cortical-cortical or cortical-subcortical neural connections which occurs as a consequence of demyelination and axonal degeneration.
The reason for this apparent discrepancy with respect to our amplitude-related findings could be due to the fact that we included MS patients with a lower EDSS in our study. Amplitude alterations seem to be more pronounced in cognitively impaired patients or when patients with high EDSS or secondary-progressive forms are included [18, 70]. It is commonly accepted, that in advanced courses of MS and in patients with a high disability, the decrease of ERP amplitude could be a consequence of widespread demyelination and axonal degeneration in the brain of these patients, causing disruption of cortical-subcortical network connections and their recruitment for neural activity [22, 23]. However, despite conservation of amplitude for standard ERP components, the use of more refined cognitive paradigms like the Posner paradigm could detect more ERP abnormalities related to silent cognitive disturbances in these types of MS patients . Thus, similar preservation of amplitude has been reported for traditional ERPs components, but specific ERP components associated to working memory paradigms showed a reduced amplitude [41, 71, 72].
Exploring the ERP latency results of our study allowed us to observe that MS patients demonstrated an increased latency in most ERP components analyzed. Thus, our results showed an increased of P2, N2 and P3 latencies in both MS groups respect to healthy subjects. Moreover, BMS patients obtained an increase of latency for N1 and P3 components in comparison with RRMS patients when standard stimuli were analysed (which involved a higher number of stimuli than target condition). Consistent with our findings on amplitude or behavioural measures, these differences seem to confirm a greater alteration in cognitive processing for MS patients with a benign profile. Increased ERP latencies in BMS as well as RRMS led us to suggest that ERPs may indicate subtle degrees of cognitive dysfunction in different courses of MS disease.
Thus, in all MS patients of our study, latencies of the ERP components P2, N2, and P3 were longer than in healthy subjects, which is quite consistent with the results of previous studies. Most studies using an ERP approach in MS have reported abnormalities of the latency of ERP components, particularly for P3, both in auditory modalities [17–21, 66–68] and visual modalities [28, 70] or using both modalities [65, 73, 74]. Some of these studies have also reported an increase in latencies or a decrease in the amplitude for N1 component [18, 28], P2 component [17–19, 21, 67, 69, 70] or N2 component [18–20, 67, 69, 70]. Similarly, a longer latency for more specific ERP components in working memory paradigms and similar tasks has been found [41, 70, 71].
In this context, these discrepancies might result from differences in the task or the degree of cognitive abnormalities in the studied patients. Thus, when results are divided in terms of different subtypes of MS, patients with a chronic stage of disease or higher neuropsychological impairment obtain more important abnormalities on ERP latencies [18, 21, 41, 65].
The current study is highly compatible with the idea that abnormalities in long-latency ERP components are not simply due to a different processing speed in sensory visual pathways, but the consequence of central nervous system demyelination. In general, the great majority of findings related to ERP latency abnormalities [23, 24] have been interpreted, either in terms of a consequence of demyelinating lesions in primary afferent pathways resulting in prolonged late ERP components [17, 18, 21], or as a functional disconnection of subcortical regions from higher cortical areas due to demyelinating plaques [22, 73, 74].
The main aim of this analysis was to determine whether there would be subtypes of MS patients based on different combinations of their cognitive profile and psychophysiological pattern, and to investigate which parameters have a greater weight to split these participants.
Taking into account the measured psychophysiological parameters of our study after performing a Posner-like task, we may conclude that from the MS patients who participated in our study, three distinct subgroups were identified (typical relapsing-remitting course and benign profile course): a "non-cognitive deficit" cluster grouped with controls, a "mild cognitive deficit" cluster with some psychophysiological abnormalities, and a "severe cognitive deficit" cluster with widespread cognitive dysfunction. Furthermore, cluster analysis revealed two more interesting findings: firstly, two control groups could be created based on their younger age as well as short ERP latencies, and secondly, some BMS patients could not be clustered into any group (outliers patients) due to their significant cognitive deficit.
Cluster 1 comprising 50% of controls was characterized by a good behavioural and psychophysiological performance with a mean Z-score lower than the other clusters in all measured domains. Participants of Cluster 2 accounted for the rest of healthy controls as well as approximately 20% of RRMS and 10% BMS patients. They were also characterized by a good behavioural performance and normal psychophysiological measures. Between these two clusters, compromising all controls and very few MS patients, the difference demonstrated by the cluster analysis solution was between the latency of P2 and P3 components, which was very close to 1.5 or 2 standard deviations from Cluster 1 in these parameters. This discrepancy between both clusters that took up all healthy subjects may be due mainly to the fact that healthy participants of Cluster 1 were relatively younger than controls of Cluster 2, although it was not statically confirmed. One hypothesis offering a psychophysiological explanation is that of age-related effects on the latencies of ERP components. There is strong evidence that P3 latency (and less so for P2) appears to be affected by age. On this basis, our findings are in line with other authors who have reported a generic increase in long-latency ERP components in controls with age [75, 76].
Most MS patients fell into the other two remaining clusters. Thus, a majority of RRMS patients were included in Cluster 3 (70% of our RRMS sample), while Cluster 4 included a proportional clustering of RRMS and BMS patients. Cluster 3 had a normal Posner task performance in terms of CRs percentage and delay in RTs, although they were also characterized by a moderate-high increase of latency-scores in analyzed ERPs (N1, P2 and P3) with respect to Cluster 1 and 2, showing a mild delay in information processing and a non-standard increase of neural activity in high-order process (LN component) associated with periods after the P3 component.
However, Cluster 4 demonstrated a global and severely altered cognitive processing, characterized by mean performances in the range of two or more standard deviations outside those of the control clusters, in all scores measured. They were characterized by a general cognitive delay, both from a behavioural perspective, with high RTs in their responses and an important increase in numbers of errors (missed), as a widespread slowing in information processing and abnormalities in high-order processing.
Finally, a specific analysis for outlier cases removed from the clustering procedure revealed that these outliers only belonged to BMS patients and they showed severe impairment in terms of behavioural performance (RTs and %CRs) and psychophysiological parameters (latencies and amplitude) in the range of four or more standard deviations from the controls. For a more detailed description of cluster features, please see the footnote in Figure 4.
Cluster analysis studies on ERP data are extremely scarce, particularly for MS patients. To our knowledge only one study has been published using cluster analysis to describe cognitive subgroups in MS, but with a neuropsychological approach . Consistent with the findings of this study, attentional and information processing deficits may be reliable, early markers that could be used to differentiate the unaffected from the mildly cognitively impaired RRMS patients.
On the other hand, there seem to be some variables that are more relevant than others for clustering the participants of our study. The new ANOVA performed with the clusters as between-subjects factors and psychophysiological scores as within-subjects factors revealed that, in general, latency parameters (N1, P2 and particularly P3 components) were the strongest scores for clustering patients. However, amplitude parameters, probably due to the higher variability within them, seem to be sensitive only to cluster splits formed mostly by patients.
The behavioural performance in our attentional task was also a sensitive variable for clustering patients, with more than 2 standard deviations in the case of RTs, and more than 3 for accuracy. Studies of attention in patients with MS have observed reduced speed of information processing, although accuracy is usually similar to that of healthy subjects, indicating that differences in accuracy between MS patients and controls are much smaller than differences in speed [4, 37, 39, 78]. Our results confirm these results at least with RRMS patients. However, when accuracy is in part determined by time required for response, differences in accuracy between MS patients and controls could be as important as differences in speed for patients with many years of disease and a benign MS profile.
Limitations and implications
Firstly, it should be noted that the cognitive abnormalities assessed in the present study using the Posner task might not be completely representative of the great variety and intensity of cognitive problems in MS. However, the main purpose of including the attentional Posner task is justified since attentional deficit and slowing of speed mental processing appears between the most impaired cognitive domains in this disease. Some variations of our experimental design may be taken into account in order to improve the sensitivity of our task, such as number, proportion and type of stimuli as inter-trial time, or response time. Further studies on new Posner version tasks may also help clarify this issue . In addition, future studies employing a standard neuropsychological assessment could help to elucidate differences between specific cognitive domains. Conventional MRI measures are also required in order to obtain a more accurate interpretation of the relationship between MRI abnormalities and neuropsychological and neurophysiological cognitive function.
Another limitation in our study certainly comes from the small sample size examined in the two subgroups of MS, particularly in the benign group. Our patients may not be fully representative of the overall group of BMS subjects, and further studies including more patient groups apart from the MS spectrum (e.g. clinically isolated syndrome, secondary-progressive or primary forms of MS) are needed to verify the exact functional role of attentional and information processing in ERPs parameters. Despite this, our findings are in line with other studies that confirm the variability of the benign course of MS. Finally, in order to assess our results from a more global point of view, it would be particularly interesting to perform the cluster analysis including a larger set of clinical and paraclinical data in order to better establish the ecological validity of a cognitive profile-based classification of MS patients, as well as the relationship between ERP abnormalities and clinical evidence.