Multiple sclerosis (MS) is an inflammatory demyelinating disease of the central nervous system with an autoimmune pathogenesis which has been the subject of thousands of clinical and experimental studies [1–3]. Although the causes triggering the autoimmune process are not entirely clear, it is generally accepted that they lie in the interplay between genetic and environmental factors [4–6]

Recently a new "actor" has appeared on the scene, namely, chronic cerebrospinal venous insufficiency (CCSVI), a condition which may be either mostly genetically determined or acquired [7–9]. CCSVI is characterised by multiple stenoses of the extracranial venous draining pathways, i.e. the internal jugular veins and the Azygous veins, which lead to collateral formation, alteration of the blood-brain barrier, and accumulation of iron, which itself could trigger and maintain the autoimmune cascade [7, 9, 10]. Some well known, but poorly explained, aspects of MS, such as the perivenular location of the lesions, the neurodegeneration and the common presence of iron deposits, seem to fit well with this hypothesis.

There is, however, no general consensus about the actual existence of CCSVI in MS, and its putative pathogenetic role, or about the most efficient tool for diagnosing it [11, 12]. Most studies finding significant differences in the prevalence of CCSVI between MS patients and controls used a venous sonographic method in accordance with Zamboni's original report [13]. However, Doepp et al. [14, 15], adopting a different sonographic approach, did not find evidence of CCSVI in a group of 56 MS patients. In two further studies, the pathogenetic role of CCSVI was challenged, as the condition was found to be linked only to more advanced disease [16], or to be present only in a very limited proportion (16%) of patients with clinically isolated syndrome [17]. Even less explored is the possibility that MS patients with CCSVI, as opposed to those without, have a characteristic disease 'phenotype'. The finding of phenotypic associations could prove particularly important, steering further pathophysiological research and making it possible to refine a putative 'vascular model' to be integrated with the existing autoimmune one.

In the present research, whose main aims were to assess the variability of CCSVI data between the centres and to seek to identify consistent relations with clinical aspects, we analysed the 'basic' clinical data of MS patients from six MS centres, five Italian and one Canadian. All the patients were submitted to venous echo-colour-Doppler examinations according to the procedure used by Zamboni et al.

The centres in which the echo-colour-Doppler examinations were performed are:

Centre 1: Department of Public Health and Neurosciences, IRCCS "C. Mondino National Institute of Neurology" Foundation, University of Pavia, Italy

Centre 2: Barrie Vascular Imaging, Barrie Ontario, Canada

Centre 3: Hospital of Civitanova, Marche, Italy

Centre 4: Department of Neurosciences, Federico II University of Naples, Italy

Centre 5: Multiple Sclerosis Centre, IRCCS Don Gnocchi, Milan, Italy

Centre 6: CCSVI Project, "Policlinico di Monza", Italy

Given the high between-centre differences both in the presence of CCSVI (found in 86.2% of patients overall, range: 74%-96%) and in the mean values of the patients' clinical parameters, the presence of possible associations between CCSVI and clinical data was tested by logistic regression analysis using a stepwise forward procedure, entering first the centre defining factor and subsequently the clinical data.

The final model proved to be adequate (chi-square (7 df) = 89.5; p < 0.001). The parameters included in the final model were (in addition to the centre defining factor) EDSS and age at onset. A higher EDSS score and an older age at onset both indicated a higher probability of CCSVI being present.

Table 6 reports the mean values of EDSS and age at disease onset (the two variables selected by logistic regression) in patients with and in those without CCSVI from each centre. Overall, the frequency of CCSVI increased with increasing EDSS values (see Figure 1: chi-square for linear association 17.6, 1 df; p < 0.001).

MS centre	CCSVI	1	2	3	4	5	6	All centres
Age at onset	+	31/112	36/200	32/105	33/87	30/40	34/66	34/610
	-	28/34	31/9	28/11	28/31	29/9	31/4	29/98
p <		ns	ns	ns	0.01	ns	Ns	0.001
EDSS	+	3.6/112	5.1/200	4.3/105	4.4/87	6.0/40	4.6/66	4.5/610
	-	3.3/36	1.7/9	2.3/11	3.6/31	4.4/9	1.5/4	3.4/98
p <		ns	0.01	ns	ns	0.01	ns	0.001

The levels of significance are also reported.

Within-centre analyses revealed the following differences: MSSS was significantly higher in CCSVI-positive patients from Centres 2,5 and 6; age at disease onset was greater in CCSVI-positive subjects from Centre 4. With regard to the different types of MS, the following percentages of patients, in the whole sample, were found to be CCSVI-positive: relapsing-remitting (RR) 82%, secondary progressive (SP) 91%, primary progressive (PP) 92% (p < 0.001). With the exception of Centre 6, progressive patients (i.e. SP+PP) were more frequently CCSVI-positive, the difference reaching statistical significance in Centres 3 and 4 (a borderline significant difference, p = 0.055, was also found in Centre 2).

In the limited literature available on this topic, CCSVI has been found in different proportions of MS subjects [20–22]. In the majority of studies, however, the frequency of CCSVI-positive patients exceeded that of CCSVI-positive healthy controls, suggesting that some link with MS (and possibly with other conditions) may exist.

However, the prevalence of CCSVI-positive patients varies greatly between centres, a fact confirmed by our data. Many factors may explain this variability in our study, including the use of different techniques, the relative unreliability of the test and the differences in the patient populations. The between-centre variability was found to be even higher when considering the single CCSVI criteria, although the third and first criteria tended to occur with greater frequency than the other three. Similar figures (even if with slightly lower percentages) were recently reported in the largest single center study by Zivadinov et al. [22]. We believe that the high frequency of the third criterion in MS patients is a particularly significant finding, since this criterion is related to direct visualisation of a venous anatomical abnormality, and should therefore be less operator-dependent. Although there were patients in whom flow abnormalities were seen in the absence of anatomical alterations, in these cases there is always a doubt that the B-mode anomalies were not seen. Since all the operators taking part in our study were trained in the method of Zamboni et al. in order to reduce inter-observer differences [23], the considerable variability in our data could be explained by the use of different equipment.

Differences in patient characteristics could account for another portion of the between-centre variability in our CCSVI findings. Indeed, higher frequencies of CCSVI were recorded in the centres which had the more severely affected patients. However, the relevance of this finding is not limited to the fact that it could partially explain the between-centre variability. Indeed, the association between CCSVI and EDSS (demonstrated by the logistic regression analysis) not only accounted for overall between-centre differences (possibly related to technical or other factors), but was also confirmed in the within-centre analyses, which showed consistently higher EDSS values in CCSVI-positive as opposed to CCSVI-negative subjects (the overall difference, corresponding to about one EDSS point, reached statistical significance in three centres). This finding, already reported by others [24], was reinforced in our study by the generally higher MSSS scores in CCSVI-positive patients.

The other variable consistently associated with CCSVI was age at disease onset, and this constituted another relevant finding: the CCSVI-positive patients were about five years older than the CCSVI-negative ones at disease onset. This association, to our knowledge never previously reported (Zivadinov et al. [22] reported only a slight difference in the frequency of CCSVI in a small group of paediatric MS), deserves particular attention. It could not be attributed only to differences in clinical course - PP subjects have generally a later disease onset and also more frequently present CCSVI - since it was also prominent in the RR subgroup. The association of older age at onset and a worse course in RR MS patients has been reported in a population-based study [25]. Moreover, a study on the clustering of multiple sclerosis recently highlighted the possible effect of 'exogenous' factors as determinants of a higher age at onset [26]. Indeed, also taking these observations into account, a putative model integrating 'autoimmune' and 'vascular' pathogenesis could fit our findings. In this framework, CCSVI could act as an 'exogenous' factor increasing mean age at onset by favouring the development of the disease in older subjects with a relatively low susceptibility to autoimmune diseases, some of whom would possibly otherwise never be ill. The recent finding that CCSVI is not linked to HLA DRB1*1501 status [27] confirms that venous insufficiency may act independently of the 'autoimmune trait'. On the contrary, these same findings are more difficult to integrate into an alternative model in which CCSVI is considered a secondary phenomenon.

Our data, while relatively robust due to the multicentre approach, need to be confirmed by replication and our hypothesis further evaluated. For instance, according to our model MS subjects without CCSVI should show a higher prevalence of MS-associated autoimmune disorders (i.e. thyroiditis or uveitis) and respond better to immunomodulating therapy.

Our data confirm that CCSVI is linked to MS, and tend to support the hypothesis that it could be a factor favouring disease development in relatively low-predisposed subjects and possibly a factor increasing disease severity. Every effort should be made to increase the reliability of the existing diagnostic techniques and to develop other ones, as the inter-centre differences observed in this and in other studies seem to be unacceptably high. The possible safety and efficacy of vascular intervention procedures should be assessed in blinded trials. Our data seem to suggest that the power of pilot interventional studies would be increased by selecting patients with a relatively older age at disease onset.

This multicentre study confirms that a high proportion MS patients have CCSVI, as assessed by venous Doppler sonography. The relative high between-centre variability, particularly of single CCSVI criteria suggests that search for more reliable diagnostic procedures should be encouraged. The finding of significant relations between CCSVI and clinical aspects of MS testifies that CCSVI and MS are related entities. Nevertheless the significance of this relation is far from being assessed and further clinical, epidemiological as well as experimental research is needed.