| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Letters to the Editor |
Department of Obstetrics and Gynecology, University of Cagliari, Ospedale San Giovanni di Dio; Via Ospedale 46, 09124 Cagliari, Italy
Editor:
We read with great interest the article by Dr Patel and colleagues in the March 1999 issue of Radiology (1). We appreciate their efforts to analyze the additional ultrasonographic (US) features of endometrioma. In their conclusion, the authors suggest that only an adnexal mass with a typical US appearance without wall nodularity and associated with hyperechoic wall foci or multilocularity is suggestive of an endometrioma. Unfortunately, Dr Patel and colleagues obtained a low agreement between test results and findings at surgery, with 
values lower than 0.6. Findings of prospective studies in the English-language literature (2–9), however, demonstrate values higher than 0.75, which indicates excellent agreement.
From the article on the retrospective study by Dr Patel and colleagues (1), we could receive the mistaken impression that US is not useful in the differential diagnosis of endometrioma; however, this is not true because these authors use a too-descriptive approach. On the contrary, only with the use of a logistic regression analysis could one evaluate the actual value of each finding. Investigators in several prospective studies (2–9) have considered the presence of homogeneous, hypoechoic tissue with low-level echoes as a major typical finding of endometrioma, as confirmed by Dr Patel and colleagues (1). Some of these authors (3–5) also introduce the presence of other minor additional findings, such as wall thickness, with evidently lower importance in comparison with the major typical findings in the diagnosis of endometrioma. These additional findings do not seem to interfere with the diagnostic accuracy, which is comparable among the different studies.
In 1992, Kupfer et al (10) (not cited by Dr Patel and colleagues) showed in a retrospective study that "the typical finding of endometrioma was the presence of a homogeneous hypoechoic ‘carpet’ of low-level echoes." This rate represents the sensitivity of the US technique, and all the prospective study findings published in subsequent years demonstrated comparable sensitivities, which ranged from 81% to 85% (2–9). These data confirm the reproducibility of the typical findings.
As also proposed by Dr Patel and colleagues (1), the use of color Doppler US could increase the diagnostic accuracy, but a fundamental criterion—the distinction between pre- and postmenopausal women—could also be used. In our opinion, examining a mixed population should be avoided because of the different performances of the test in pre- and postmenopausal patients. Color Doppler US could be used to reduce the number of false-negative findings due to a typical appearance that is associated with a wall nodularity.
In 1998, our group proposed a new approach (9), which is based on the localization of vessels and intensity of arterial flow, with the use of color Doppler energy (power Doppler) US imaging. This allows for the inclusion of atypical endometriomas in which no flow is detected in the echogenic portion because of the presence of a clot that can be differentiated from an intracystic vegetation. When color Doppler energy US imaging findings are positive and the result of the CA-125 test is higher than 25 U/mL, the probability of an endometrioma is high (95.6%) (9). As determined by using logistic regression, the absence of the US and biochemical factors reduces the possibility of the presence of an ovarian endometrioma to 1.4%.
Footnotes
Recent eLetters to the Editor are available at http://radiology.rsnajnls.org. eLetters that are no longer posted under "Recent eLetters" can be found as a link in the related article or by browsing through past Tables of Contents.
References
Department of Radiology, Mayo Clinic Scottsdale, 13400 East Shea Boulevard, Scottsdale, AZ 85259*; Department of Radiology, University of California, San Francisco
We appreciate the interest of Dr Guerriero and colleagues in our article (1) and the opportunity their letter provides to further discuss our results. Unfortunately, these authors believe our article suggests that only those masses exhibiting low-level internal echoes, no neoplastic features, and hyperechoic wall foci or multilocularity have an appearance suggestive of an endometrioma. This is a mistaken impression that may have arisen from the conclusion statement in the abstract, in which we indicate that this combination of gray-scale US features results in the highest positive likelihood ratio for the diagnosis of endometrioma. By focusing only on the combination of features that results in the highest positive likelihood ratio for the diagnosis of endometrioma, our critics suggest that our data are not in agreement with those in earlier published reports (2–9). This is not the case.
In our study, the feature combination of "low-level internal echoes and no neoplastic features" had a sensitivity of 75%, specificity of 91%, and positive likelihood ratio of 8 for the diagnosis of endometrioma (1); this matches values in Dr Guerriero and colleagues' own study, in which they found a sensitivity of 83%, specificity of 93%, and positive likelihood ratio of 12 for the diagnosis of an adnexal mass as an endometrioma by using similar criteria (6).
Investigators in studies with a higher diagnostic performance of US for endometrioma use patient selection criteria that decrease potential false-positive results. Thus, Dr Guerriero and colleagues found that US had a positive likelihood ratio of 28 for endometrioma in a population of women with a persistent adnexal mass (7). In our study, if spontaneously resolving lesions are excluded from consideration (1, table 2), the feature combination of "low-level internal echoes and no neoplastic features" had a positive likelihood ratio of 27. Indeed, our data regarding the diagnostic performance of US with this feature combination are very concordant with those of previous studies. One may then erroneously begin to question the additional value of our study.
We believe our study represents a valuable additional contribution to the literature regarding the US diagnosis of endometrioma for several reasons. First, as noted by Dr Guerriero and colleagues in their letter, previous studies have added additional requirements to the definition of a "classic" endometrioma beyond the presence of low-level echoes and absence of neoplastic features, such as a thin wall, thick wall, or visualization within the ovary, which our study indicates are not helpful features.
Second, we recognize that there is a receiver operating characteristic (ROC) curve with respect to the diagnosis of endometrioma; this has been ignored in previous investigations. If one chooses to have a higher specificity in the diagnosis of endometrioma, one must sacrifice sensitivity. For example, our study findings showed that if one seeks to avoid the misdiagnosis of an acute hemorrhagic cyst as an endometrioma by using the feature combination of low-level internal echoes, no neoplastic features, and absence of features of acute hemorrhage as the diagnostic criteria, the sensitivity for identifying endometriomas decreases from 75% to 65%, the specificity increases from 91% to 96%, and the positive likelihood ratio increases from 8 to 17.
Third, we identified two additional features that serve to distinguish endometriomas from hemorrhagic ovarian cysts: the presence of hyperechoic wall foci and multilocularity. Thus, if one chooses to achieve the highest positive likelihood ratio for the diagnosis of endometrioma by using these additional criteria, which serve to eliminate those hemorrhagic cysts that account for false-positive findings, the sensitivity for identifying endometriomas further decreases to 45%, while the specificity increases to 99% and the positive likelihood ratio increases to 48.
In their letter, Dr Guerriero and colleagues raise a distinction in the use of US to diagnose endometriomas between pre- and postmenopausal women. We agree, and our study findings help to clarify this difference. To our knowledge, no one has shown that there is a difference in the US appearance of endometriomas between pre- and postmenopausal women; what differs is the pretest probability of the lesions that can mimic certain US feature combinations seen with endometriomas. Thus, for premenopausal patients, we advocate the use of stricter diagnostic criteria, sacrificing sensitivity for specificity, before one concludes that a mass is an endometrioma, because self-limiting hemorrhagic cysts are common in premenopausal women. Since we strive to avoid unnecessary surgery in premenopausal women by employing US follow-up for those lesions that do not demonstrate neoplastic features, we require the additional criterion of the presence of hyperechoic wall foci or multilocularity or documented persistence before concluding that a mass is an endometrioma. Other factors that we did not study, such as the size of the lesion, may also be relevant.
For postmenopausal women, in whom hemorrhagic ovarian cysts are not expected to occur, we would choose a different ROC threshold in concluding that a mass is most likely an endometrioma. For these women, the combination of low-level internal echoes and no neoplastic features is sufficiently diagnostic to allow surgery to proceed, with the expectation that one is dealing with an endometrioma. Our study findings showed that benign cystic teratomas can demonstrate low-level internal echoes and no neoplastic features. Furthermore, we know from anecdotal experience that benign mucinous cystadenomas also can demonstrate these features. However, since the surgical work-up and approach are similar for endometriomas and benign cystic neoplasms in the postmenopausal patient, these potential false-positive findings are not clinically important, and we can afford to choose this lower ROC threshold. This contrasts sharply with the clinically meaningful false-positive designation of a self-limiting hemorrhagic ovarian cyst as an endometrioma in the premenopausal woman, which could result in unnecessary surgery.
As we suggested in our discussion, color Doppler US may prove useful to evaluate wall nodularity. In those masses in which we cannot be certain that neoplastic features are absent because of the presence of wall nodularity, color Doppler US may be able to help differentiate between (a) avascular wall nodularity in endometriomas due to fibrosis and clot deposition and (b) wall nodularity in neoplasms in which vascularity may be apparent. We applaud Dr Guerriero and colleagues for their efforts in investigating the utility of color Doppler US imaging in this regard (9).
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
