Article Text

Role of transoesophageal echocardiography in infective endocarditis
  1. Med Klinik II
  2. Universität Erlangen-Nürnberg
  3. Östl.Stadtmauerstr. 29
  4. D-91054 Erlangen, Germany
  5. email: werner.g.daniel{at}

Statistics from

Infective endocarditis continues to be an underdiagnosed and undertreated disease, in spite of a mortality which exceeds that of acute myocardial infarction. In the diagnosis of this disease, echocardiography—including transoesophageal echocardiography—has been a major breakthrough. While the traditional hallmarks of endocarditis were fever together with a new or changed heart murmur, cutaneous signs, and embolic events, echocardiography has enabled the direct visualisation of endocarditic lesions. Indeed, the widely acknowledged Duke criteria for endocarditis1 list echocardiographic signs as a major diagnostic criterion, which together with positive blood cultures allows the definite diagnosis of endocarditis (table 1). Recent work indicates that echocardiography should have an even more prominent place, because patients nowadays are very frequently pretreated with antibiotics before any blood cultures are drawn, leading to a high incidence of culture negative endocarditis.2

Table 1

The Duke criteria for endocarditis

Transthoracic echocardiography detects vegetations with approximately 70% sensitivity.3 4 The use of transoesophageal echocardiography increases sensitivity and specificity to about 90%,4-7 mainly by detecting small vegetations that escape the transthoracic diagnosis. The diagnostic advantage of transoesophageal echocardiography is even more pronounced in abscesses and prosthetic valves, where transthoracic echocardiography may miss half or more of the lesions.8 9 At the same time, extension of the disease to other valves or to surrounding tissue, valvar defects, chordal rupture, fistulas, and other complications can be diagnosed, the severity of valvar regurgitation and hence hemodynamic stability can be assessed, and size and mobility of vegetations provide a rough estimate of the likelihood of embolism.5 10 11 Management decisions, especially the decision to proceed to early surgery, are facilitated by findings such as large vegetations, severe valvar regurgitation, abscess formation, prosthetic valve dehiscence, etc. Conversely, a negative transoesophageal echocardiogram has a negative predictive power of over 90%.12 The negative predictive power of a repeated negative transoesophageal echocardiogram after a few days is very high (over 95%), and is the most convincing clinical way to exclude an endocarditis if there is a clinical suspicion of this disease.

In view of this diagnostic potential, should every febrile patient get an echocardiogram, a transoesophageal echocardiogram, or possibly more than one? Apart from the practicality, pretest likelihood considerations indicate that this is not the case. While fever is indeed a very common condition, endocarditis is not a frequent disease. Estimates put the incidence at approximately 14–19 per million persons per year.13 14 There should therefore be a reasonable degree of suspicion (or pretest likelihood) before sending a patient with fever to the echocardiography laboratory. In practice this means that: (1) other common causes for fever should be clinically unlikely; or (2) the patient should be at risk for endocarditis, such as patients with degenerative valve disease, prosthetic heart valves, immunosuppression, etc; or (3) there should be typical other signs of endocarditis, such as a new heart murmur, cutaneous signs, or embolism.

However, given one or more of the above, there should be no hesitation to proceed from transthoracic to transoesophageal echocardiography if the former is inconclusive. In fact, in most cases even in the presence of a positive transthoracic echocardiogram, an additional transoesophageal study is helpful to reveal the full extent of the disease, especially in the presence of a prosthetic heart valve, in endocarditis not responding promptly to antibiotics, or other complications.

If the suspicion of endocarditis persists and the transoesophageal echocardiogram is negative, the question of a repeat study arises. The best time interval between studies has never been evaluated systematically. Although large vegetations take at least several days to grow, staphylococcal destruction of valves can occur very rapidly. The exact timeframe therefore must be individualised, weighing the risk of overlooking severe disease by postponing a repeat study for too long against the risk of ending up with another false negative or inconclusive study by performing it too early. If the clinical suspicion is rather low, repetition of the study after a week or so seems reasonable.

Have the features of endocarditis changed during recent years ?

The retrospective analysis of Netzer and colleagues in this issue15 could not address systematically the utility of echocardiography in general, and transoesophageal echocardiography in particular, owing to its retrospective nature, going back to times before transoesophageal echocardiography was available. It is a chart review of 212 patients hospitalised at their institution with “definite” or “possible” endocarditis according to the Duke criteria, comparing characteristics of the disease in the time intervals 1980-87 and 1988-95. Their findings are mostly, though not completely, in line with our current understanding of this disease.

There are good reasons to believe that the face of infective endocarditis has been changing in recent decades. Presumed factors for this change are:

  • the rise of previously absent risk conditions such as prosthetic valves, immunosuppression, intravenous drug abuse, long term intensive care, and others;

  • an aging population with a higher incidence of degenerative valve disease;

  • the decline in rheumatic valve disease;

  • changes in the spectrum of causative organisms, in particular a shift from Streptococcus viridans to Staphylococcus aureus.

The most frequent causative organisms in both time intervals in the present study were streptococci, although there was a mild increase in staphylococcal endocarditis over time. This is in line with other reports,13 although in recent times staphylococcal disease, with its dismal prognosis, is clearly on the rise.16 Mortality in this series was relatively low (15%), which is confirmed by recent estimates from comparable populations (20% in the 1980s,13 16% in 1996-816).

Factors strongly associated with a higher mortality were, among others, age, “neurologic signs” on admission or during the disease course, “fatigue”, identification of an abscess on echocardiography, and higher levels of C reactive protein. These findings are consistent with former reports and not surprising. In contrast, previous studies found a notably worse prognosis for prosthetic valve endocarditis,3 13 possibly owing to the more aggressive causative bacteria (mostly staphylococci). In the present study, however, prosthetic valve endocarditis was not found to be associated with higher mortality. Nevertheless, the high mortality of staphylococcal endocarditis (19%) was confirmed.

Interestingly, the authors report that in their intravenous drug addicts streptococci were the most frequent causative agents, in strong contrast to previous studies. The authors explain this difference by the Swiss programme of improving hygiene in addicts by education and distribution of sterile needles.

Hence, has infective endocarditis essentially remained the same disease as 20 years ago? That is unlikely. For an adequate interpretation of the present findings, the following possible confounding issues should be kept in mind.

Referral bias—This is not a population based study, and it is unclear how representative for the disease spectrum the patients of this tertiary referral centre were during the two time intervals. It is also not clear whether patients referred directly from outside physicians to cardiac surgery were included in the analysis.

Multiple comparisons and sample size—Many of the differences in table 3 (Factors associated with mortality) in the paper by Netzer and colleagues15 are only marginally significant and the level of significance was not corrected for multiple comparisons. Therefore, it is very possible that they do not represent meaningful associations. On the other hand, some differences may be not significant because of small absolute numbers, but still are noteworthy, such as the doubling or tripling of incidence of drug addiction and HIV as underlying conditions. Further, although the paper states a lack of association between mortality and prosthetic valve endocarditis, it elsewhere acknowledges substantially higher mortality in both early (21%) and late (24%) prosthetic valve endocarditis, compared to a native valve endocarditis mortality of 14%. Clearly, the lack of significance here is misleading and should not be construed as evidence against an underlying difference in mortalities.

Case definitions—The authors used the Duke criteria (“possible” or “definite”) for diagnosis of infective endocarditis, which are widely accepted. However, the Duke criteria1 do not specify minimum requirements for possible endocarditis. “Possible endocarditis” according to this classification implies “findings consistent with infective endocarditis that fall short of ‘definite’, but not ‘rejected’”. The initiators of this classification in their original paper state about patients classified as “possible endocarditis”: “This category includes a wide range of indeterminate cases that have very different probabilities of infective endocarditis” and “classification . . . into the category of possible endocarditis does not imply that antimicrobial therapy should or should not be given.” It is therefore difficult to ascertain which level of confidence that diagnosis of endocarditis carries in this study. In this regard it is also puzzling to see that not only in the early 1980s, but also in the more recent time periods, in every fifth case the diseased valve could not be identified.

Finally, proving that a difference does not exist between two sets of observations is always harder than finding a “significant” difference. Therefore, the absence of significant differences between the two time intervals does not imply rigorously that there was no such difference. Infective endocarditis, a notoriously protean disease, may in fact be changing its appearance under our noses.


View Abstract

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.