Jump to comment:

• Intracardiac thrombi in stroke: common, but how relevant?
  Sumeet Singhal
  Published on: 4 November 2004

• Published on: 4 November 2004

  Intracardiac thrombi in stroke: common, but how relevant?

  • Sumeet Singhal, Neurology SpR
  • Other Contributors:
    • Nottingham

  Dear Editor,

  The study by Sen et al in last month's JNNP (2004;75:1421-1425) regarding risk factors for intracardiac thrombi post-stroke is interesting, but the results require cautious interpretation. The authors present data on a topic of great relevance to clinical management of stroke patients, namely which patients should be referred for transoesophageal echocardiography (TOE). Their attempt to deriv...

  Show More

  Dear Editor,

  The study by Sen et al in last month's JNNP (2004;75:1421-1425) regarding risk factors for intracardiac thrombi post-stroke is interesting, but the results require cautious interpretation. The authors present data on a topic of great relevance to clinical management of stroke patients, namely which patients should be referred for transoesophageal echocardiography (TOE). Their attempt to derive clinical criteria that may assist in selecting patients for TOE is welcome. The study was well-designed, being prospective in nature and studying 151 consecutive patients with stroke/TIA, minimising selection bias. TOE analysis was well-blinded and showed good interobserver reliability, suggesting that TOE interpretation by cardiologists is reproducible. Results are clearly presented. However, I have some reservations regarding the interpretation and application of their results.

  Firstly, the classification of stroke types is confusing. The authors imply that ‘large strokes’ means ‘large vessel strokes’, which presumably includes both large and small infarct sizes, but this is not stated. The failure to define the radiological criteria for these terms makes it difficult to know how to apply their findings in practice. There are also discrepancies between the results for ‘large’ and ‘small’ (vessel) strokes, based on radiological criteria, and for TACS/PACS/POCS and LACS based on clinical criteria. The claim that ‘stroke location’, based on these clinical criteria, is not associated with intracardiac thrombi (ICT), appears at odds with the finding that ‘large’ (vessel) stroke is associated with ICT. Thus, someone with a clinical TACS, confirmed on CT, would be considered at high risk for ICT based on radiological but not clinical grounds. Should they undergo TOE or not?

  Statistical interpretation of study may have been flawed. A p value of 0.08 for the univariate association between symptomatic coronary artery disease (sCAD) and ICT would not conventionally lead one to believe there “might” be a significant relationship, as the authors suggest. Nor would the subsequent p value on multivariate analysis of 0.017 be generally considered “highly significant”, a term usually reserved for p values < 0.01. The conversion of a non-significant result on univariate analysis to a significant one on multivariate analysis is counter-intuitive, and suggests the result was a spurious finding. It is just possible that negative relationships existed between other covariates and ICT. Hence, only when these covariates were controlled for (on multivariate analysis), would the result reach significance, but this is not borne out by their findings on univariate analysis of other variables. Similarly, the relationship between ischaemic ECG changes and ICT is only significant (p value not given) on multivariate analysis. These two associations are therefore of dubious significance.

  The association between atrial fibrillation (AF) and ICT was at best weak (p=0.045), particularly considering multiple comparisons were performed. Furthermore, the association was lost on multivariate analysis, suggesting there was no independent relationship, the opposite of what is claimed in the first sentence of the conclusions. Even if this relationship holds (which would be biologically plausible), it is hard to see how the presence of ICT in a stroke patient with AF would result in a “possible requirement for more intensive anticoagulation”. The likely mechanism by which AF increases the risk of stroke is through cardioembolism, hence the rationale for anticoagulating stroke patients with AF in the first place.

  Only two relationships apparently reached statistical significance on both univariate and multivariate analyses, those of ‘large’ (vessel) stroke with ICT, and of left ventricular dysfunction with ICT. Problems with the former concept have been discussed. Regarding the latter, a p value was not given. Judging by the confidence intervals, which are very close to 1.0 at the lower end, the p values were probably not far off 0.05, which is at best borderline in the face of multiple comparisons. Furthermore, this particular parameter was measured on TOE itself, making it a nonsensical criterion by which to determine the need for TOE. Their assertion that transthoracic echocardiography (TTE) could be used to select candidates for TOE based on LV systolic function may well be valid, but assumes that the same patients would have LV dysfunction picked up on TTE as on TOE, for the relationship with ICT to hold.

  The suggestion that “patients presenting with lacunar stroke should not necessarily be excluded from TOE to rule out a cardiac source of embolism”, because of the finding that 28% of patients with LACS had ICT, highlights the crux of this debate. Are we interested in factors predicting a cardioembolic mechanism, when presence of ICT might then justifiably lead to anticoagulation, or are we interested in the presence of ICT per se? Theoretically, it is implausible that lacunar stroke is likely to be influenced by ICT, as the authors themselves point out. Furthermore, no association between either LACS or ‘small’ (vessel) stroke and ICT was found, unlike with ‘large’ (vessel) strokes. Thus, one could equally interpret the 28% prevalence of ICT in patients with LACS as suggesting that ICT is a common incidental finding on multiplanar TOE of no pathological relevance. Indeed, the frequency of asymptomatic ICT is unknown. It would have been more useful to assess if presence of ICT on TOE accurately predicts cardioembolism as a mechanism for stroke, as judged by multiterritorial infarcts, recent MI, maximal deficit at onset of the event, and absence of another likely stroke mechanism (such as severe carotid stenosis or lacunar stroke on the background of hypertension), all factors that are considered to be suggestive of a cardiac embolism [Ref]. To this end, it may be helpful to stratify risk of cardioembolic stroke based on ICT size and location. If a predictive value of ICT could be shown, it would lend some credence to their assertion that “the use of these variables [sCAD, ECG evidence of ischaemia and large stroke on CT/MRI] to focus TOE investigation on those with a high probability of harbouring an intracradiac thrombus may improve outcome”.

  From the data presented, I would conclude:
  1) ICT detected by multiplanar TOE post-stroke are probably more common than previously suspected;
  2) there is as yet unsubstantiated evidence that certain clinical variables may increase the chance of detecting ICT post-stroke;
  3) the clinical relevance of these ICT are uncertain;
  4) an intervention trial, comparing outcome from anticoagulation with that from standard antiplatelet therapy, for unselected stroke patients with ICT on TOE is badly needed.

  Until a causal association between all detected ICT and risk of any type stroke can be determined, it appears inappropriate to use clinical variables, whether or not they are confirmed to be associated with presence of ICT, to select patients for TOE, if the aim is to anticoagulate positive cases. For the present, the decision to use TOE should continue to be based on clinical evidence of a cardioembolic mechanism for the event, not on the likelihood of finding an intracardiac thrombus.

  Reference

  Kelley RE, Minagar A. Cardioembolic stroke: an update. South Med J. 2003;96:343-9.

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top