- Email: [email protected]
The positive and negative predictive value of transabdominal color Doppler ultrasound for diagnosing ovarian torsion in pediatric patients
Journal of Pediatric Surgery (2013) 48, 1283–1287
www.elsevier.com/locate/jpedsurg
The positive and negative predictive value of transabdominal color Doppler ultrasound for diagnosing ovarian torsion in pediatric patients Jessica A. Naiditch, Katherine A. Barsness ⁎ Division of Pediatric Surgery, Ann & Robert H. Lurie Children's Hospital of Chicago, Chicago, IL, USA Northwestern University Feinberg School of Medicine, Chicago, IL, USA Received 11 February 2013; accepted 8 March 2013
Key words: Gonadal torsion; Ovarian torsion; Transabdominal ultrasound; False positive
Abstract Purpose: The purposes of this study were to (1) determine the positive and negative predictive value (NPV) of transabdominal color Doppler ultrasound (CDU) for diagnosing ovarian torsion (OT) in pediatric patients and 2) identify predictors of a false-positive CDU result for OT. Methods: An IRB-approved retrospective chart review was performed on all female patients who underwent transabdominal CDU evaluation of the ovaries (664 CDUs in 605 patients) for acute abdominal pain. CDU reports were categorized as positive for OT if the report stated “cannot rule out torsion” or “positive for torsion.” Results: There were 47 false-positive ultrasounds, 3 false negatives, 11 true positives, and 603 true negatives for OT. Sensitivity was 78.6%, specificity 92.3%, positive predictive value (PPV) 19.0%, and NPV 99.5%. False-positive CDU when compared to true positives were more common in older patients (p = 0.004) and were more commonly read as “cannot rule out torsion” (p b 0.001). Ovarian cysts were larger in true-positive CDU than in false-positive CDU (p b 0.001). However, cyst presence/absence did not predict a true positive result. Conclusion: Transabdominal CDU has a low PPV and a high NPV for ovarian torsion in pediatric patients. False-positive results are more common in older patients and associated with small ovarian cysts. © 2013 Elsevier Inc. All rights reserved.
Ovarian torsion (OT) is an uncommon but important cause of acute abdominal pain in the female pediatric population. Early diagnosis of OT is critical for preventing necrosis of the ovary and fallopian tube [1–3]. Transabdom⁎ Corresponding author. Ann & Robert H. Lurie Children's Hospital of Chicago, 225 East Chicago Avenue, Box 63, Chicago, IL 60611, USA. Tel.: +1 312 227 4735; fax: +1 312 227 9678. E-mail address: [email protected] (K.A. Barsness). 0022-3468/$ – see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jpedsurg.2013.03.024
inal color Doppler ultrasound imaging (CDU) with blood flow analysis is the imaging modality of choice for OT as it can be performed quickly, provides adequate visualization of the ovaries, and requires no ionizing radiation. However, CDU findings in OT can be non-specific and associated with a low sensitivity and specificity for diagnosis OT in children and adolescents [4–6]. These non-specific findings often lead to an equivocal study with the inability to radiologically rule out torsion. Clarifying the positive and negative
1284
J.A. Naiditch, K.A. Barsness
predictive value for transabdominal CDU may aid surgeons in their decision process in determining which patients, for whom ovarian torsion is in the differential diagnosis, should undergo exploration. The current pediatric literature on the value of CDU for diagnosing OT is primarily based on patients who had OT confirmed at surgical exploration [5]. In order to determine a positive predictive value for transabdominal CDU, an accurate analysis also requires the true-negative and falsepositive CDUs for patients who did not undergo surgical exploration. The purposes of this study were to (1) determine the positive and negative predictive value of transabdominal CDU in diagnosing OT in pediatric patients and (2) identify predictors of a false-positive CDU result for OT.
1. Methods and subjects After obtaining approval from the internal review board (IRB approval no. 2011-14728), we performed a retrospective chart review for all patients who underwent emergent CDU evaluation of the ovaries due to acute abdominal symptoms at a free-standing children's hospital from January 2007 to June 2011. Patients were identified through the use of Xenobase, allowing ultrasound reports to be interrogated for the terms related to ovarian torsion. These terms included “ovary,” “ovaries,” “ovarian,” and “torsion.” For patients with these key terms, charts were reviewed. Patients being evaluated with CDU evaluation of the ovaries for acute abdominal symptoms were included in the study. In order to validate our search methodology, all patients with the International Classification of Diseases Ninth Edition Clinical Modification (ICD-9-CM) for ovarian torsion (620.5) who had undergone CDU at our institution were
Fig. 1
confirmed as included in our initial patient search. Deidentified data were collected from these charts including age at evaluation, presenting signs and symptoms, laboratory values, CDU findings including the presence or absence of flow in the ovaries as well as any abnormal anatomic findings. The 12 pediatric radiologists who read CDUs during the study period follow standard radiologic criteria for the diagnosis of ovarian torsion. This includes some combination of a unilateral enlarged ovary (either based on contralateral ovarian size or standard published ovarian volumes for age), free pelvic fluid, lack of arterial or venous flow, or a twisted vascular pedicle. For the purposes of this study, CDU reports were categorized as positive for torsion if the final report stated “cannot rule out torsion” or “positive for torsion,” as torsion remains in the differential for all of these patients following CDU. Studies were considered negative for torsion if the report stated “negative for torsion” or did not comment on torsion. Attending radiologists confirm all CDU results during the day. CDUs are initially read by the radiology resident on-call at night. For patients with a CDU read as positive for torsion or “cannot rule out torsion” during the evening hours, the attending radiologist is available to confirm the study read that night. Patient disposition whether they were discharged to home, admitted to the hospital for observation, or underwent emergent or urgent operation within 12 h of CDU and operative findings were recorded. When data were available, body mass index-for-age percentile (BMI-FAP) was calculated for children 2 years or older and weight-for-age percentile (weight-FAP) for children less than 2 years old using Epi Info (Centers for Disease Control, Atlanta, GA). Patient factors and outcomes of interest were compared using chi-squared test for categorical data and ANOVA and Student's t-test for continuous data with significance determined by p b 0.05.
Therapeutic pathway for female pediatric patients evaluated by CDU for acute abdominal symptoms.
PPV and NPV of transabdominal CDU
1285
Table 1 CDU results and presence of ovarian torsion: all cases, n = 664. CDU result for OT
Ovarian torsion present Sensitivity Specificity PPV a NPV b a b
Yes No 78.6% 92.3% 19.0% 99.5%
Positive
Negative
11 47
3 603
PPV positive predictive value. NPV negative predictive value.
2. Results Six hundred sixty-four CDU studies were performed on 605 female pediatric patients evaluated for acute abdominal pain (Fig. 1). For patients who underwent N 1 CDU, CDUs were confirmed as performed N 30 days apart and therefore considered unique episodes of acute abdominal pain. There were 6 studies read as positive, 52 read as “cannot rule out torsion,” and 606 studies read as negative or that did not mention torsion. The positive and “cannot rule out” torsion studies were combined and considered as positive studies for OT. Assuming all patients with torsion underwent operation, there were 47 false-positive ultrasounds, 3 false negatives, 11 true positives, and 603 true negatives for ovarian torsion (Table 1). Sensitivity was 78.6%, specificity 92.3%, positive predictive value (PPV) 19.0%, and negative predictive value (NPV) 99.5%. This assumes that there were no missed diagnoses of OT among all patients who underwent CDU for acute abdominal pain. As it is possible that the diagnosis was missed, we also evaluated CDU results only for patients that were surgically explored for any indication (n = 113). When only patients surgically explored for any indication were considered, there were 21 false-positive ultrasounds, 3 false negatives, 11 true positives, and 78 true negatives for ovarian torsion (Table 2). For these operative cases the sensitivity was 78.6%, specificity 78.8%, PPV 34.4%, and NPV 96.3%.
Table 2 CDU results and presence of ovarian torsion: operatively confirmed cases, n = 113. CDU result for OT
Ovarian Torsion Present Sensitivity Specificity PPV a NPV b a b
Yes No 78.6% 78.8% 34.4% 96.3%
PPV positive predictive value. NPV negative predictive value.
Positive
Negative
11 21
3 78
Eleven of 24 patients taken to the operating room due to concern for OT and with a CDU “positive” for ovarian torsion did not have torsion (45.8%). In order to determine if there were differentiating factors that could predict a false-positive CDU, we compared patient factors and CDU findings for false-positive and true-positive CDU results (Table 3). When comparing all false-positive cases with true-positive cases, false-positive CDU occurred in older patients (14.9 ± 3.5 years vs. 11.4 ± 3.1 years; p = 0.004), often had normal venous blood flow (72.3% vs. 36.4%, p = 0.022) and were more commonly read as “cannot rule out torsion” (97.8% vs. 54.6%; p b 0.001). The presence of an ovarian cyst alone was not associated with true positive. However, ovarian cysts when present were larger in true-positive CDU (10.4 ± 3.8 cm) than in false-positive CDU (3.7 ± 1.7 cm; p b 0.001). When comparing true-positive CDU with only operatively confirmed false-positive CDU, the only finding that differed from the above comparison is that patients were of a similar age group (13.7 ± 4.6 vs. 11.4 ± 3.2, p = 0.15).
3. Discussion An expeditious diagnosis of ovarian torsion is imperative for preventing ovarian necrosis and tissue loss. Transvaginal US, perhaps the most sensitive and specific imaging study for OT, is inappropriate in the majority of the pediatric patient population [7,8]. The most accurate imaging study available for female pediatric patients presenting with acute abdominal symptoms is transabdominal color Doppler ultrasound; however; transabdominal CDU findings for OT can be nonspecific [5,9–11]. The CDU features associated with OT are highly variable and no single indicator has been found to have a high specificity. Complete arterial obstruction probably does not occur in most cases [10]. Doppler signal is sometimes difficult to obtain even in normal ovaries using transabdominal CDU [11]. Some have said that visualization of multiple follicles in the outer portion of a unilaterally enlarged ovary is specific for OT, which is found in 74% of torsion cases [9]. Others have suggested that the most specific finding for OT is measuring a unilateral enlarged ovary [5]. In the present study, we demonstrate that transabdominal CDU has a positive predictive value of 19.0% to 34.4% and a negative predictive value of 96.3% to 99.5% for diagnosing of OT in pediatric patients. The 7% to 18.6% false-positive rate of CDU for OT is associated with a high rate of negative surgical explorations for OT, which occurred in 11 of 24 explorations for OT, 45.8%. Comparison of true-positive and false-positive CDU suggests that false-positive CDU may occur in older patients and patients who more often have normal venous blood flow on CDU. When an ovarian cyst is present, falsepositive CDU is associated with smaller ovarian cysts. Obesity was not associated with a false-positive CDU. The literature is lacking in studies evaluating the utility of transabdominal CDU for diagnosis OT in pediatric patients. Transabdominal gray-scale US for the diagnosis of OT in
1286 Table 3
J.A. Naiditch, K.A. Barsness Comparison of false positive and true positive transabdominal color Doppler ultrasound findings.
Age (years), avg ± st dev Obese US findings Free fluid Blood flow a Venous Normal Diminished Absent Arterial Normal Diminished Absent Cyst in ovary in question Largest diameter (cm), avg ± st dev Mass Largest diameter (cm), avg ± st dev Final US read Cannot rule out torsion Positive for torsion
True positive (N = 11)
False positive, all (N = 47)
p
False positive, operative (N = 21)
11.4 ± 3.1 4 (36.4%)
14.9 ± 3.5 16 (34.0%)
0.004 0.84
13.7 ± 4.6 3 (14.3%)
0.15 0.32
4 (36.4%)
17 (36.2%)
0.74
7 (33.3%)
0.82
4 (36.4%) 2 (18.2%) 5 (45.5%)
34 (72.3%) 9 (19.2%) 4 (8.5%)
5 (45.5%) 2 (18.2%) 4 (36.4%) 7 (63.6%) 10.4 ± 3.8 0
33 (70.2%) 10 (21.3%) 4 (8.5%) 23 (48.9%) 3.7 ± 1.7 2 (4.3%) 3.3 ± 1.4
6 (54.6%) 5 (45.5%)
46 (97.8%) 1 (2.1%)
0.022
p
0.020 13 (61.9%) 7 (33.3%) 1 (4.8%)
0.12
0.58 b 0.001 0.82
0.06 13 (61.9%) 7 (33.3%) 1 (4.8%) 8 (38.1%) 3.5 ± 1.3 2 (9.5%) 3.3 ± 1.4
b 0.001
0.32 b 0.001 0.78 0.020
20 (95.2%) 1 (4.8%)
The bold indicates statistical significance. a For the ovary read as positive for torsion.
children has previously been studied [5,12]. Chang et al. [12] reviewed 49 pediatric cases of OT for which 43 patients underwent gray-scale US imaging. Of these, 41 suggested ovarian pathology and 3 torsions were identified. Oltmann et al. [5] retrospectively reviewed 328 ovarian operations, 97 of which were for torsion. In their study, transabdominal US had a sensitivity of 51% for OT. The higher sensitivity found in the study presented here may be attributed to the addition of color Doppler to the US. Others have questioned the usefulness of spectral Doppler flow in assessing for OT, as normal venous and arterial flow has been demonstrated in pathologically confirmed necrotic ovaries [13,14]. The continued low sensitivity of transabdominal US with the addition of color Doppler flow assessment highlights that OT is still a clinical diagnosis that cannot be dependent on imaging findings. For the first time in the literature, we define the positive and negative predictive value of CDU for OT in children. Previous work evaluating the sensitivity and specificity of transabdominal US for OT was derived only from patients operatively confirmed to have OT. The exclusion of non-operative patients prevented the determination of positive and negative predictive values of CDU for OT in pediatric patients. The high negative predictive value of CDU defined in this study suggests that surgeons may reliably rule out torsion in the setting of a negative CDU for OT. However, in the setting of a CDU read as positive for torsion, the low positive predictive value of CDU for OT requires surgeons to rely on their clinical judgment as to whether OT is highly suspected. Previous work has suggested that obesity may decrease the sensitivity of US for ovarian pathology. Yoo et al. [15]
evaluated the ability of US to delineate ovarian structures in obese adolescents of PCOS and found that, when compared to a control group of normal-weight adolescents, US was less accurate in the obese PCOS group. Although they did not study OT specifically, their work suggests that the sensitivity of transabdominal CDU for OT may be impaired in the obese child. To the contrary, obesity was not associated with falsepositive CDU results in the current study. With a high rate of false-positive studies, we sought to identify markers for a false-positive CDU. When compared to patients with true-positive studies, false-positive studies occurred in older patients. Although false-positive studies did not correlate with obesity, the larger body habitus of the older female child may affect the ability to perform a reliable transabdominal ultrasound study. The presence of a cyst or mass alone did not predict a true-positive result. However, patients with true-positive studies had larger cysts when present. Not surprisingly, a CDU that “cannot rule out torsion” was highly associated with a false-positive study. For clinical purposes, CDU studies read as “cannot rule out torsion” are treated as positive studies and considered within the clinical context of individual patients. For this reason, “cannot rule out torsion” CDUs were included in the positive study group and not surprisingly are associated with false-positive studies. We acknowledge the limitations inherent in the retrospective nature of this study. Patients discharged after a negative CDU may have presented to a different institution with recurrent pain and subsequently diagnosed with OT, causing an overestimate in true negatives. In order to attempt to overcome this pitfall, we performed two analyses as
PPV and NPV of transabdominal CDU above, the second only considering patients operatively confirmed to not have OT as true negatives. This study is dependent on the pediatric radiologists that read the CDU studies and our results may be affected by institutional biases in what constitutes a positive or negative study. In conclusion, CDU has a high negative predictive value and therefore offers utility in ruling out ovarian torsion in female pediatric patients presenting with acute abdominal pain. However, a CDU read as positive for torsion or that cannot rule out torsion must be considered within the clinical context of each patient with consideration for diagnostic laparoscopy when the clinical picture is not clear. Depending on CDU for a definitive diagnosis of OT may result in many negative surgical explorations and therefore should not be the only indication for operative intervention.
Acknowledgments Thank you to Andrea Chen who diligently collected data for this project.
References [1] Kokoska ER, Keller MS, Weber TR. Acute ovarian torsion in children. Am J Surg 2000;180:462-5. [2] Mordehai J, Mares AJ, Barki Y, et al. Torsion of uterine adnexa in neonates and children: a report of 20 cases. J Pediatr Surg 1991;26:1195-9. [3] Spigland N, Ducharme JC, Yazbeck S. Adnexal torsion in children. J Pediatr Surg 1989;24:974-6. [4] Graif M, Shalev J, Strauss S, et al. Torsion of the ovary: sonographic features. AJR Am J Roentgenol 1984;143:1331-4. [5] Oltmann SC, Fischer A, Barber R, et al. Cannot exclude torsion—a 15year review.J Pediatr Surg 2009;44:1212-6 discussion 1217. [6] Warner MA, Fleischer AC, Edell SL, et al. Uterine adnexal torsion: sonographic findings. Radiology 1985;154:773-5. [7] Fleischer AC, Kepple DM. Transvaginal color duplex sonography: clinical potentials and limitations. Semin Ultrasound CT MR 1992;13: 69-80. [8] Van Voorhis BJ, Schwaiger J, Syrop CH, et al. Early diagnosis of ovarian torsion by color Doppler ultrasonography. Fertil Steril 1992;58:215-7. [9] Graif M, Itzchak Y. Sonographic evaluation of ovarian torsion in childhood and adolescence. AJR Am J Roentgenol 1988;150:647-9. [10] Oelsner G, Bider D, Goldenberg M, et al. Long-term follow-up of the twisted ischemic adnexa managed by detorsion. Fertil Steril 1993;60: 976-9. [11] Surratt JT, Siegel MJ. Imaging of pediatric ovarian masses. Radiographics 1991;11:533-48. [12] Chang YJ, Yan DC, Kong MS, et al. Adnexal torsion in children. Pediatr Emerg Care 2008;24:534-7. [13] Hurh PJ, Meyer JS, Shaaban A. Ultrasound of a torsed ovary: characteristic gray-scale appearance despite normal arterial and venous flow on Doppler. Pediatr Radiol 2002;32:586-8. [14] Servaes S, Zurakowski D, Laufer MR, et al. Sonographic findings of ovarian torsion in children. Pediatr Radiol 2007;37:446-51. [15] Yoo RY, Sirlin CB, Gottschalk M, et al. Ovarian imaging by magnetic resonance in obese adolescent girls with polycystic ovary syndrome: a pilot study. Fertil Steril 2005;84:985-95.
1287
Discussion The positive and negative predictive value of TransAbdominal Color Doppler Ultrasound for the Diagnosis of Ovarian Torsion in Pediatric Patients Presented by: Jessica Naiditch, MD, Chicago, IL Discussant: unidentified: Let me just ask you about the methodology because I know that all of the patients who had acute abdominal pain had a color Doppler. Is that right? Response: DR. NAIDITCH: The way that the study was done, we identified people through the ultrasound results. It is not that every patient with acute abdominal pain that came into our hospital had a color Doppler ultrasound. It's that we took all of the patients that had transabdominal color Doppler ultrasounds and then determined which of those were being evaluated for acute abdominal pain. That's how the study was designed, from the ultrasound forward. Unidentified discussant: Okay, but among those 600-some patients, did any of them end up having appendicitis or other diagnosis that would make it irrelevant to look at the ovaries? Response: DR. NAIDITCH: Yes, there were a number of patients in whom the indication for the ultrasound was lower abdominal pain, so the ovaries and the appendix were both evaluated. That's why in the analysis when we talk about which patients went to surgery, it's all patients that went for any operative indication. Unidentified discussant: I just want to say this is a great study. My question is about the 26 patients that had a positive ultrasound but ended up in the observed arm, who did you use to make that criterion because if a patient came in with abdominal pain and they had a positive ultrasound for torsion, I think most would consider at least laparoscopy just to make sure. Response: DR. NAIDITCH: As I mentioned in the beginning, those with positive studies were those that either were truly read as positive or unable to rule out torsion. When we have those kinds of patients in our emergency room, we realize that many of these patients are not going to end up having torsion based on our previous experience. So it really comes down to clinical acumen and level of suspicion. Yes, if they continued to have pain and you're not sure, laparoscopy is definitely advisable to ensure that torsion is not present. But the false positives happen often enough that we don't necessarily consider that alone an indication for exploration.
www.elsevier.com/locate/jpedsurg
The positive and negative predictive value of transabdominal color Doppler ultrasound for diagnosing ovarian torsion in pediatric patients Jessica A. Naiditch, Katherine A. Barsness ⁎ Division of Pediatric Surgery, Ann & Robert H. Lurie Children's Hospital of Chicago, Chicago, IL, USA Northwestern University Feinberg School of Medicine, Chicago, IL, USA Received 11 February 2013; accepted 8 March 2013
Key words: Gonadal torsion; Ovarian torsion; Transabdominal ultrasound; False positive
Abstract Purpose: The purposes of this study were to (1) determine the positive and negative predictive value (NPV) of transabdominal color Doppler ultrasound (CDU) for diagnosing ovarian torsion (OT) in pediatric patients and 2) identify predictors of a false-positive CDU result for OT. Methods: An IRB-approved retrospective chart review was performed on all female patients who underwent transabdominal CDU evaluation of the ovaries (664 CDUs in 605 patients) for acute abdominal pain. CDU reports were categorized as positive for OT if the report stated “cannot rule out torsion” or “positive for torsion.” Results: There were 47 false-positive ultrasounds, 3 false negatives, 11 true positives, and 603 true negatives for OT. Sensitivity was 78.6%, specificity 92.3%, positive predictive value (PPV) 19.0%, and NPV 99.5%. False-positive CDU when compared to true positives were more common in older patients (p = 0.004) and were more commonly read as “cannot rule out torsion” (p b 0.001). Ovarian cysts were larger in true-positive CDU than in false-positive CDU (p b 0.001). However, cyst presence/absence did not predict a true positive result. Conclusion: Transabdominal CDU has a low PPV and a high NPV for ovarian torsion in pediatric patients. False-positive results are more common in older patients and associated with small ovarian cysts. © 2013 Elsevier Inc. All rights reserved.
Ovarian torsion (OT) is an uncommon but important cause of acute abdominal pain in the female pediatric population. Early diagnosis of OT is critical for preventing necrosis of the ovary and fallopian tube [1–3]. Transabdom⁎ Corresponding author. Ann & Robert H. Lurie Children's Hospital of Chicago, 225 East Chicago Avenue, Box 63, Chicago, IL 60611, USA. Tel.: +1 312 227 4735; fax: +1 312 227 9678. E-mail address: [email protected] (K.A. Barsness). 0022-3468/$ – see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jpedsurg.2013.03.024
inal color Doppler ultrasound imaging (CDU) with blood flow analysis is the imaging modality of choice for OT as it can be performed quickly, provides adequate visualization of the ovaries, and requires no ionizing radiation. However, CDU findings in OT can be non-specific and associated with a low sensitivity and specificity for diagnosis OT in children and adolescents [4–6]. These non-specific findings often lead to an equivocal study with the inability to radiologically rule out torsion. Clarifying the positive and negative
1284
J.A. Naiditch, K.A. Barsness
predictive value for transabdominal CDU may aid surgeons in their decision process in determining which patients, for whom ovarian torsion is in the differential diagnosis, should undergo exploration. The current pediatric literature on the value of CDU for diagnosing OT is primarily based on patients who had OT confirmed at surgical exploration [5]. In order to determine a positive predictive value for transabdominal CDU, an accurate analysis also requires the true-negative and falsepositive CDUs for patients who did not undergo surgical exploration. The purposes of this study were to (1) determine the positive and negative predictive value of transabdominal CDU in diagnosing OT in pediatric patients and (2) identify predictors of a false-positive CDU result for OT.
1. Methods and subjects After obtaining approval from the internal review board (IRB approval no. 2011-14728), we performed a retrospective chart review for all patients who underwent emergent CDU evaluation of the ovaries due to acute abdominal symptoms at a free-standing children's hospital from January 2007 to June 2011. Patients were identified through the use of Xenobase, allowing ultrasound reports to be interrogated for the terms related to ovarian torsion. These terms included “ovary,” “ovaries,” “ovarian,” and “torsion.” For patients with these key terms, charts were reviewed. Patients being evaluated with CDU evaluation of the ovaries for acute abdominal symptoms were included in the study. In order to validate our search methodology, all patients with the International Classification of Diseases Ninth Edition Clinical Modification (ICD-9-CM) for ovarian torsion (620.5) who had undergone CDU at our institution were
Fig. 1
confirmed as included in our initial patient search. Deidentified data were collected from these charts including age at evaluation, presenting signs and symptoms, laboratory values, CDU findings including the presence or absence of flow in the ovaries as well as any abnormal anatomic findings. The 12 pediatric radiologists who read CDUs during the study period follow standard radiologic criteria for the diagnosis of ovarian torsion. This includes some combination of a unilateral enlarged ovary (either based on contralateral ovarian size or standard published ovarian volumes for age), free pelvic fluid, lack of arterial or venous flow, or a twisted vascular pedicle. For the purposes of this study, CDU reports were categorized as positive for torsion if the final report stated “cannot rule out torsion” or “positive for torsion,” as torsion remains in the differential for all of these patients following CDU. Studies were considered negative for torsion if the report stated “negative for torsion” or did not comment on torsion. Attending radiologists confirm all CDU results during the day. CDUs are initially read by the radiology resident on-call at night. For patients with a CDU read as positive for torsion or “cannot rule out torsion” during the evening hours, the attending radiologist is available to confirm the study read that night. Patient disposition whether they were discharged to home, admitted to the hospital for observation, or underwent emergent or urgent operation within 12 h of CDU and operative findings were recorded. When data were available, body mass index-for-age percentile (BMI-FAP) was calculated for children 2 years or older and weight-for-age percentile (weight-FAP) for children less than 2 years old using Epi Info (Centers for Disease Control, Atlanta, GA). Patient factors and outcomes of interest were compared using chi-squared test for categorical data and ANOVA and Student's t-test for continuous data with significance determined by p b 0.05.
Therapeutic pathway for female pediatric patients evaluated by CDU for acute abdominal symptoms.
PPV and NPV of transabdominal CDU
1285
Table 1 CDU results and presence of ovarian torsion: all cases, n = 664. CDU result for OT
Ovarian torsion present Sensitivity Specificity PPV a NPV b a b
Yes No 78.6% 92.3% 19.0% 99.5%
Positive
Negative
11 47
3 603
PPV positive predictive value. NPV negative predictive value.
2. Results Six hundred sixty-four CDU studies were performed on 605 female pediatric patients evaluated for acute abdominal pain (Fig. 1). For patients who underwent N 1 CDU, CDUs were confirmed as performed N 30 days apart and therefore considered unique episodes of acute abdominal pain. There were 6 studies read as positive, 52 read as “cannot rule out torsion,” and 606 studies read as negative or that did not mention torsion. The positive and “cannot rule out” torsion studies were combined and considered as positive studies for OT. Assuming all patients with torsion underwent operation, there were 47 false-positive ultrasounds, 3 false negatives, 11 true positives, and 603 true negatives for ovarian torsion (Table 1). Sensitivity was 78.6%, specificity 92.3%, positive predictive value (PPV) 19.0%, and negative predictive value (NPV) 99.5%. This assumes that there were no missed diagnoses of OT among all patients who underwent CDU for acute abdominal pain. As it is possible that the diagnosis was missed, we also evaluated CDU results only for patients that were surgically explored for any indication (n = 113). When only patients surgically explored for any indication were considered, there were 21 false-positive ultrasounds, 3 false negatives, 11 true positives, and 78 true negatives for ovarian torsion (Table 2). For these operative cases the sensitivity was 78.6%, specificity 78.8%, PPV 34.4%, and NPV 96.3%.
Table 2 CDU results and presence of ovarian torsion: operatively confirmed cases, n = 113. CDU result for OT
Ovarian Torsion Present Sensitivity Specificity PPV a NPV b a b
Yes No 78.6% 78.8% 34.4% 96.3%
PPV positive predictive value. NPV negative predictive value.
Positive
Negative
11 21
3 78
Eleven of 24 patients taken to the operating room due to concern for OT and with a CDU “positive” for ovarian torsion did not have torsion (45.8%). In order to determine if there were differentiating factors that could predict a false-positive CDU, we compared patient factors and CDU findings for false-positive and true-positive CDU results (Table 3). When comparing all false-positive cases with true-positive cases, false-positive CDU occurred in older patients (14.9 ± 3.5 years vs. 11.4 ± 3.1 years; p = 0.004), often had normal venous blood flow (72.3% vs. 36.4%, p = 0.022) and were more commonly read as “cannot rule out torsion” (97.8% vs. 54.6%; p b 0.001). The presence of an ovarian cyst alone was not associated with true positive. However, ovarian cysts when present were larger in true-positive CDU (10.4 ± 3.8 cm) than in false-positive CDU (3.7 ± 1.7 cm; p b 0.001). When comparing true-positive CDU with only operatively confirmed false-positive CDU, the only finding that differed from the above comparison is that patients were of a similar age group (13.7 ± 4.6 vs. 11.4 ± 3.2, p = 0.15).
3. Discussion An expeditious diagnosis of ovarian torsion is imperative for preventing ovarian necrosis and tissue loss. Transvaginal US, perhaps the most sensitive and specific imaging study for OT, is inappropriate in the majority of the pediatric patient population [7,8]. The most accurate imaging study available for female pediatric patients presenting with acute abdominal symptoms is transabdominal color Doppler ultrasound; however; transabdominal CDU findings for OT can be nonspecific [5,9–11]. The CDU features associated with OT are highly variable and no single indicator has been found to have a high specificity. Complete arterial obstruction probably does not occur in most cases [10]. Doppler signal is sometimes difficult to obtain even in normal ovaries using transabdominal CDU [11]. Some have said that visualization of multiple follicles in the outer portion of a unilaterally enlarged ovary is specific for OT, which is found in 74% of torsion cases [9]. Others have suggested that the most specific finding for OT is measuring a unilateral enlarged ovary [5]. In the present study, we demonstrate that transabdominal CDU has a positive predictive value of 19.0% to 34.4% and a negative predictive value of 96.3% to 99.5% for diagnosing of OT in pediatric patients. The 7% to 18.6% false-positive rate of CDU for OT is associated with a high rate of negative surgical explorations for OT, which occurred in 11 of 24 explorations for OT, 45.8%. Comparison of true-positive and false-positive CDU suggests that false-positive CDU may occur in older patients and patients who more often have normal venous blood flow on CDU. When an ovarian cyst is present, falsepositive CDU is associated with smaller ovarian cysts. Obesity was not associated with a false-positive CDU. The literature is lacking in studies evaluating the utility of transabdominal CDU for diagnosis OT in pediatric patients. Transabdominal gray-scale US for the diagnosis of OT in
1286 Table 3
J.A. Naiditch, K.A. Barsness Comparison of false positive and true positive transabdominal color Doppler ultrasound findings.
Age (years), avg ± st dev Obese US findings Free fluid Blood flow a Venous Normal Diminished Absent Arterial Normal Diminished Absent Cyst in ovary in question Largest diameter (cm), avg ± st dev Mass Largest diameter (cm), avg ± st dev Final US read Cannot rule out torsion Positive for torsion
True positive (N = 11)
False positive, all (N = 47)
p
False positive, operative (N = 21)
11.4 ± 3.1 4 (36.4%)
14.9 ± 3.5 16 (34.0%)
0.004 0.84
13.7 ± 4.6 3 (14.3%)
0.15 0.32
4 (36.4%)
17 (36.2%)
0.74
7 (33.3%)
0.82
4 (36.4%) 2 (18.2%) 5 (45.5%)
34 (72.3%) 9 (19.2%) 4 (8.5%)
5 (45.5%) 2 (18.2%) 4 (36.4%) 7 (63.6%) 10.4 ± 3.8 0
33 (70.2%) 10 (21.3%) 4 (8.5%) 23 (48.9%) 3.7 ± 1.7 2 (4.3%) 3.3 ± 1.4
6 (54.6%) 5 (45.5%)
46 (97.8%) 1 (2.1%)
0.022
p
0.020 13 (61.9%) 7 (33.3%) 1 (4.8%)
0.12
0.58 b 0.001 0.82
0.06 13 (61.9%) 7 (33.3%) 1 (4.8%) 8 (38.1%) 3.5 ± 1.3 2 (9.5%) 3.3 ± 1.4
b 0.001
0.32 b 0.001 0.78 0.020
20 (95.2%) 1 (4.8%)
The bold indicates statistical significance. a For the ovary read as positive for torsion.
children has previously been studied [5,12]. Chang et al. [12] reviewed 49 pediatric cases of OT for which 43 patients underwent gray-scale US imaging. Of these, 41 suggested ovarian pathology and 3 torsions were identified. Oltmann et al. [5] retrospectively reviewed 328 ovarian operations, 97 of which were for torsion. In their study, transabdominal US had a sensitivity of 51% for OT. The higher sensitivity found in the study presented here may be attributed to the addition of color Doppler to the US. Others have questioned the usefulness of spectral Doppler flow in assessing for OT, as normal venous and arterial flow has been demonstrated in pathologically confirmed necrotic ovaries [13,14]. The continued low sensitivity of transabdominal US with the addition of color Doppler flow assessment highlights that OT is still a clinical diagnosis that cannot be dependent on imaging findings. For the first time in the literature, we define the positive and negative predictive value of CDU for OT in children. Previous work evaluating the sensitivity and specificity of transabdominal US for OT was derived only from patients operatively confirmed to have OT. The exclusion of non-operative patients prevented the determination of positive and negative predictive values of CDU for OT in pediatric patients. The high negative predictive value of CDU defined in this study suggests that surgeons may reliably rule out torsion in the setting of a negative CDU for OT. However, in the setting of a CDU read as positive for torsion, the low positive predictive value of CDU for OT requires surgeons to rely on their clinical judgment as to whether OT is highly suspected. Previous work has suggested that obesity may decrease the sensitivity of US for ovarian pathology. Yoo et al. [15]
evaluated the ability of US to delineate ovarian structures in obese adolescents of PCOS and found that, when compared to a control group of normal-weight adolescents, US was less accurate in the obese PCOS group. Although they did not study OT specifically, their work suggests that the sensitivity of transabdominal CDU for OT may be impaired in the obese child. To the contrary, obesity was not associated with falsepositive CDU results in the current study. With a high rate of false-positive studies, we sought to identify markers for a false-positive CDU. When compared to patients with true-positive studies, false-positive studies occurred in older patients. Although false-positive studies did not correlate with obesity, the larger body habitus of the older female child may affect the ability to perform a reliable transabdominal ultrasound study. The presence of a cyst or mass alone did not predict a true-positive result. However, patients with true-positive studies had larger cysts when present. Not surprisingly, a CDU that “cannot rule out torsion” was highly associated with a false-positive study. For clinical purposes, CDU studies read as “cannot rule out torsion” are treated as positive studies and considered within the clinical context of individual patients. For this reason, “cannot rule out torsion” CDUs were included in the positive study group and not surprisingly are associated with false-positive studies. We acknowledge the limitations inherent in the retrospective nature of this study. Patients discharged after a negative CDU may have presented to a different institution with recurrent pain and subsequently diagnosed with OT, causing an overestimate in true negatives. In order to attempt to overcome this pitfall, we performed two analyses as
PPV and NPV of transabdominal CDU above, the second only considering patients operatively confirmed to not have OT as true negatives. This study is dependent on the pediatric radiologists that read the CDU studies and our results may be affected by institutional biases in what constitutes a positive or negative study. In conclusion, CDU has a high negative predictive value and therefore offers utility in ruling out ovarian torsion in female pediatric patients presenting with acute abdominal pain. However, a CDU read as positive for torsion or that cannot rule out torsion must be considered within the clinical context of each patient with consideration for diagnostic laparoscopy when the clinical picture is not clear. Depending on CDU for a definitive diagnosis of OT may result in many negative surgical explorations and therefore should not be the only indication for operative intervention.
Acknowledgments Thank you to Andrea Chen who diligently collected data for this project.
References [1] Kokoska ER, Keller MS, Weber TR. Acute ovarian torsion in children. Am J Surg 2000;180:462-5. [2] Mordehai J, Mares AJ, Barki Y, et al. Torsion of uterine adnexa in neonates and children: a report of 20 cases. J Pediatr Surg 1991;26:1195-9. [3] Spigland N, Ducharme JC, Yazbeck S. Adnexal torsion in children. J Pediatr Surg 1989;24:974-6. [4] Graif M, Shalev J, Strauss S, et al. Torsion of the ovary: sonographic features. AJR Am J Roentgenol 1984;143:1331-4. [5] Oltmann SC, Fischer A, Barber R, et al. Cannot exclude torsion—a 15year review.J Pediatr Surg 2009;44:1212-6 discussion 1217. [6] Warner MA, Fleischer AC, Edell SL, et al. Uterine adnexal torsion: sonographic findings. Radiology 1985;154:773-5. [7] Fleischer AC, Kepple DM. Transvaginal color duplex sonography: clinical potentials and limitations. Semin Ultrasound CT MR 1992;13: 69-80. [8] Van Voorhis BJ, Schwaiger J, Syrop CH, et al. Early diagnosis of ovarian torsion by color Doppler ultrasonography. Fertil Steril 1992;58:215-7. [9] Graif M, Itzchak Y. Sonographic evaluation of ovarian torsion in childhood and adolescence. AJR Am J Roentgenol 1988;150:647-9. [10] Oelsner G, Bider D, Goldenberg M, et al. Long-term follow-up of the twisted ischemic adnexa managed by detorsion. Fertil Steril 1993;60: 976-9. [11] Surratt JT, Siegel MJ. Imaging of pediatric ovarian masses. Radiographics 1991;11:533-48. [12] Chang YJ, Yan DC, Kong MS, et al. Adnexal torsion in children. Pediatr Emerg Care 2008;24:534-7. [13] Hurh PJ, Meyer JS, Shaaban A. Ultrasound of a torsed ovary: characteristic gray-scale appearance despite normal arterial and venous flow on Doppler. Pediatr Radiol 2002;32:586-8. [14] Servaes S, Zurakowski D, Laufer MR, et al. Sonographic findings of ovarian torsion in children. Pediatr Radiol 2007;37:446-51. [15] Yoo RY, Sirlin CB, Gottschalk M, et al. Ovarian imaging by magnetic resonance in obese adolescent girls with polycystic ovary syndrome: a pilot study. Fertil Steril 2005;84:985-95.
1287
Discussion The positive and negative predictive value of TransAbdominal Color Doppler Ultrasound for the Diagnosis of Ovarian Torsion in Pediatric Patients Presented by: Jessica Naiditch, MD, Chicago, IL Discussant: unidentified: Let me just ask you about the methodology because I know that all of the patients who had acute abdominal pain had a color Doppler. Is that right? Response: DR. NAIDITCH: The way that the study was done, we identified people through the ultrasound results. It is not that every patient with acute abdominal pain that came into our hospital had a color Doppler ultrasound. It's that we took all of the patients that had transabdominal color Doppler ultrasounds and then determined which of those were being evaluated for acute abdominal pain. That's how the study was designed, from the ultrasound forward. Unidentified discussant: Okay, but among those 600-some patients, did any of them end up having appendicitis or other diagnosis that would make it irrelevant to look at the ovaries? Response: DR. NAIDITCH: Yes, there were a number of patients in whom the indication for the ultrasound was lower abdominal pain, so the ovaries and the appendix were both evaluated. That's why in the analysis when we talk about which patients went to surgery, it's all patients that went for any operative indication. Unidentified discussant: I just want to say this is a great study. My question is about the 26 patients that had a positive ultrasound but ended up in the observed arm, who did you use to make that criterion because if a patient came in with abdominal pain and they had a positive ultrasound for torsion, I think most would consider at least laparoscopy just to make sure. Response: DR. NAIDITCH: As I mentioned in the beginning, those with positive studies were those that either were truly read as positive or unable to rule out torsion. When we have those kinds of patients in our emergency room, we realize that many of these patients are not going to end up having torsion based on our previous experience. So it really comes down to clinical acumen and level of suspicion. Yes, if they continued to have pain and you're not sure, laparoscopy is definitely advisable to ensure that torsion is not present. But the false positives happen often enough that we don't necessarily consider that alone an indication for exploration.