- Email: [email protected]
Can we preoperatively risk stratify ovarian masses for malignancy?
Journal of Pediatric Surgery (2010) 45, 130–134
www.elsevier.com/locate/jpedsurg
Can we preoperatively risk stratify ovarian masses for malignancy? Sarah C. Oltmann a , Nilda Garcia a , Robert Barber b , Rong Huang b , Barry Hicks a , Anne Fischer a,⁎ a
University of Texas Southwestern Medical Center, Dallas, TX 75235, USA Children's Medical Center, Dallas, TX 75235, USA
b
Received 30 September 2009; accepted 6 October 2009
Key words: Ovarian mass; Ovarian malignancy; Pediatric surgery; Gynecology
Abstract Purpose: Given a 10% malignancy rate in pediatric ovarian masses, what preoperative factors are helpful in distinguishing those at higher risk to risk stratify accordingly? Methods: After institutional review board approval (IRB#022008-095), a 15½-year retrospective review of operative ovarian cases was performed. Results: A total of 424 patients were identified, with a mean age 12.5 years (range, 1 day to 19 years), without an age disparity between benign (12.54 years, 89%) and malignant (11.8 years, 11%) cases. The 1- to 8-year age group had the highest percentage of malignancies (22%; odds ratio [OR], 3.02; 95% confidence interval [CI], 1.33-6.86). A chief complaint of mass or precocious puberty versus one of pain had an OR for malignancy of 4.84 and 5.67, respectively (95% CI, 2.48-9.45 and 1.60-20.30). Imaging of benign neoplasms had a mean size of 8 cm (range, 0.9-36 cm) compared with malignancies at 17.3 cm (6.2-50 cm, P b .001). An ovarian mass size of 8 cm or longer on preoperative imaging had an OR of 19.0 for malignancy (95% CI, 4.42-81.69). Ultrasound or computed tomographic findings of a solid mass, although infrequent, were most commonly associated with malignancy (33%-60%), compared with reads of heterogeneous (15%-21%) or cystic (4%-5%) lesions. The malignancies (n = 46) included germ cell (50%, n = 23), stromal (28%, n = 13), epithelial (17%, n = 8), and other (4%, n = 2). Tumor markers obtained in 71% of malignancies were elevated in only 54%, whereas 6.5% of those sent in benign cases were similarly elevated. Elevated beta-human chorionic gonadotropin (β-HCG), alpha fetoprotein (αFP), and cancer antigen 125 (CA-125) were significantly associated with malignancy (P b .02) and an elevated carcinoembryonic antigen (CEA) was not (P = .1880). Conclusion: This reported series of pediatric ovarian masses demonstrates that preoperative indicators that best predict an ovarian malignancy are a complaint of a mass or precocious puberty, a mass exceeding 8 cm or a mass with solid imaging characteristics. Those patients aged 1 to 8 years have the greatest incidence of malignancy. Tumor markers, positive or negative, were not conclusive in all cases but useful for postoperative surveillance. © 2010 Elsevier Inc. All rights reserved.
Presented at the 40th Annual Meeting of the American Pediatric Surgical Association, Fajardo, Puerto Rico, May 28-June 1, 2009. ⁎ Corresponding author. Tel.: +1 214 456 6040; fax: +1 214 456 6320. E-mail address: [email protected] (A. Fischer). 0022-3468/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.jpedsurg.2009.10.022
Although ovarian malignancy is relatively rare in the pediatric population constituting only 1% of all childhood malignancies, ovarian masses in childhood do have a 10%
Stratified ovarian masses and malignancy malignancy rate [1,2]. Strategies for ovarian salvage have been slow to be implemented given the unclear malignancy risk of a given ovarian mass upon presentation; high oophorectomy rates for nonmalignancies have continued to be reported despite the endorsement of ovarian salvage in the recent literature [1,3]. It is unclear how best to determine preoperatively which subset of patients with an ovarian lesion is at higher risk for malignancy, particularly because these factors may impact the decision to perform an oophorectomy at the time of surgery. Significant overlap in the clinical presentation exists between benign and malignant pathology, and dogma surrounding the risk of malignancy in ovarian masses has little evidence-based support within the current literature. In an effort to better ascertain preoperative indicators of malignancy compared with benign masses, we analyzed our experience with pediatric ovarian pathology at a major academic center to see if patient characteristics, laboratory values, and radiologic parameters exist, which can identify those patients most at risk for ovarian malignancy.
1. Methods After institutional review board approval (IRB#-022008095), International Classification of Diseases-9th revision (ICD-9) codes and pathology database were used to identify those patients from birth to 19 years of age surgically managed for ovarian findings at a major academic center consisting of a large, free-standing children's hospital and a major metropolitan county hospital from January 1, 1993 to June 30, 2008. Those patients with tubo-ovarian abscesses or isolated ectopic pregnancy were excluded. A retrospective review of admission notes, radiology reports, operative reports, laboratory results, and pathology reports was performed by a single author for internal consistency. Descriptive analyses were performed using frequency distributions, means, and SEM. χ2 and Fisher's Exact tests were used to compare proportions between 2 groups. MantelHaenszel method was used to calculate relative risk and its confidence intervals. Two-sample t test was used to compare means between the 2 groups. The significance level was set as P value less than .05. GraphPad Prism 5.01 and SAS 9.1 were used for statistical calculation.
131 Table 1 Demographics of malignant versus benign pediatric ovarian masses
No. of Patients No. of Cases Age (y), mean ± SEM Chief complaint Mass Pain Precocious puberty Prenatal imaging Incidental finding Per Orem (PO) intolerance Other/not documented WBC N12,000 Tumor markers percent abnormal findings (abnormal/drawn) beta-human chorionic gonadotriopin (β-HCG) a alpha fetoprotein (αFP) cancer antigen 125 (CA-125), carcinoembryonic antigen (CEA) Mean size by imaging (cm)
Malignant
Benign
46 (11%) 48 (11%) 11.8 ± 0.64
378 (89%) 390 (89%) 12.54 ± 0.26
22 (46%) 20 (42%) 4 (8%) 0 0 0
58 (15%) 255 (65%) 9 (2%) 11 (3%) 35 (9%) 3 (1%)
2 (4%)
19 (5%)
7 (15%) 54% (19/35)
59 (15%) 6.5% (8/122)
.92 b.001
22% (6/27)
0% (0/107)
b.001
52% (13/25) 71% (5/7) 25% (2/8) 17.3 (6.2-50)
1.9% (1/54) 22% (7/32) 5% (1/20) 8 (0.9-36)
b.001
.64 b.001 .001 .02
.83
b.02 .188 b.001
WBC indicates white blood cell. a Those patients who were pregnant at the time of evaluation were excluded from this subset.
P b .01; Fig. 1) in comparison to 10% seen in older girls. This finding was further substantiated by a 3-fold greater odds ratio for malignancy in the 1- to 8-year group compared with the 15- to 19-year olds (Table 2).
2. Results A total of 424 patients were identified, who underwent a total of 438 operative cases with ovarian findings. Forty-six patients (11%) were found to have malignant disease, with 2 requiring reoperation for local recurrence. The mean (SEM) age of those patients with malignancy was 11.8 (±0.64) years, similar to the benign population (12.54 ± 0.26 years; Table 1). However, when looking at the relative percentage of malignancy by age intervals, the subset aged 1 to 8 years had the highest percentage of malignant neoplasms (22%,
P
Fig. 1
Pathology distribution by age group.
132 Table 2
S.C. Oltmann et al. Table 3
Odds ratios for malignancy Odds ratio
Chief complaint Mass 4.84 Precocious puberty 5.67 Pain 1.0 Age at presentation (y) b1 0.364 1-8 3.02 9-14 1.338 15-19 1.0 Imaging characteristic (Ultrasound) Heterogeneous 3.62 Cystic 1.0 Solid 10.13 Imaging characteristic (CT) Heterogeneous 6.84 Cystic 1.0 Solid 39.0 Mass size by imaging (cm) ≥5 18.9 ≥6 31.4 ≥8 19.0
Malignancy by cell type
95% confidence interval 2.48-9.45 1.60-20.03
0.046-2.853 1.33-6.86 0.664-2.695
1.15-11.36 1.83-56.05 1.49-31.36
n Germ cell Yolk sac Immature teratoma Mixed germ cell Dysgerminoma Sex cord stromal Juvenile granulosa cell Sertoli-Leydig Indeterminant sex cord Epithelial Borderline serous tumor Carcinoma Other Lymphoma Leukemia
23 (50%) 4 5 9 5 13 (28%) 8 4 1 8 (17%) 6 2 2 (4%) 1 1
3.99-381 1.14-314.6 1.90-520.27 4.42-81.69
The presenting symptoms for those patients with malignancy were equally divided between a mass (n = 22, 46%) and pain (n = 20, 42%), whereas precocious puberty (n = 4, 8%) accounted for the rest (Table 1). However, because pain was also the primary complaint in the majority (65%) of the benign population, pain as a symptom was not a distinguishing characteristic. In contrast, those patients with either a chief complaint of mass or precocious puberty were significantly more likely to have a malignancy than those patients with a complaint of pain (Table 2). On laboratory evaluation, a leukocytosis exceeding 12,000 was found with equal incidence in both benign and malignant cases (Table 1). Although elevated tumor markers (β-HCG, αFP, CA-125) were highly associated with malignancy (P b .02), they were elevated only 54% of the time. Thus, in a comparable number of incidences (46%), the tumor markers were actually normal in the setting of an ovarian malignancy and surprisingly elevated in 6.5% of benign lesions. Elevated CEA, commonly seen in malignancies in adults, was less commonly assessed or detected in any of the malignancies (in 2 of 8 cases) and was not found to be associated with malignancy (P = .188). Patients were most commonly imaged with ultrasound and/or computed tomographic (CT) scan. Thirty patients had ultrasound preoperatively, 33 patients had CT scan, and 4 underwent magnetic resonance imaging evaluation. Malignant lesions all exceeded 6 cm, with an average size of 17.3 cm (range, 6.2-50 cm), significantly larger than the benign lesions averaging 8 cm (range, 0.9-36 cm; P b .001). We statistically analyzed 3 different diameters to determine which would be the most helpful clinically: 5 cm, the traditional cutoff; 6 cm, the size that all
malignancies exceeded in our series; and 8 cm [3,4]. Although the odds ratio is greatest for those diameters of 6 cm or wider at 31.4, the wide 95% confidence interval of 1.90 to 520.27 is less precise. Thus, the ≥ 8-cm-or-more cutoff was not only significant with an odds ratio of 19.0, but also had a narrower and thus more accurate confidence interval (Table 2). In addition to evaluating the mean lesion size by imaging, the nature of the lesion or mass as described by the radiologist is also important. Forty-three percent of solid lesions (6 of 14) imaged by ultrasound or CT scan were associated with malignancy, whereas complex/heterogeneous and cystic masses were each associated with malignancy in only 18% (30 of 171) and 4% (6 of 139) of cases, respectively. Resultant odds ratios for malignancy in ovarian masses that were solid or heterogeneous masses by CT were 39 and 6.48, respectively, in comparison to a cystic lesion (Table 2). Likewise, similar odds ratios were seen upon analysis of ultrasound characteristics. Because treatment and prognosis can vary significantly by malignant cell type, the pathology was also noted (Table 3).
Fig. 2
Distribution of malignancy cell type by age group.
Stratified ovarian masses and malignancy Germ cell tumors comprised 50% of all the pediatric ovarian malignancies; sex cord stromal, 28%; epithelial, 17%; and other, 4%. This distribution changed with patient age (Fig. 2). Germ cell tumors accounted for most ovarian malignancies in all age groups and the greatest in those 9 to 14 years. Sex cord stromal tumors comprised 25% to 33% of malignancies in all groups over the age of 1. Epithelial tumors only occurred in those patients over the age of 12 years and presented more often as the less aggressive serous borderline tumor than the typical epithelial carcinoma seen in adults. One patient was found to have a recurrence of acute lymphoblastic leukemia in the ovary, and another patient had primary lymphoma of the ovary.
3. Discussion Ovarian malignancies in the pediatric population are rare. In a population-based study, age-adjusted incidence for malignant pediatric ovarian tumors varied 10-fold for each age interval from only 0.102 per 100,000/y in those girls 9 years or younger, 1.072 per 100,000/y in those age 10 to 19 years, to 11.446 per 100,000/y in the adult female population [5]. Although those patients 10 years and older may have a higher age-adjusted incidence in the populationbased study, this does not take into consideration the increased incidence of other benign ovarian lesions seen within this age group. Our study, which determined percentage of malignancies in the context of an ovarian mass, found the greatest percentage of malignancies in those patients aged 1 to 8 years. As demonstrated with our pathology distribution by age in Fig. 1, perimenarchal patients tend to have a greater number of benign lesions by way of cysts and neoplasms, whereas those patients hormonally quiescent are less likely to present with an ovarian mass at all. But when they do, malignancy may be present up to a quarter of the time. As a result, a higher index of suspicion for malignancy is needed in the patient aged 1 to 8 years, as indicated by a 3-fold increase in the odds ratio. Conversely, it is exceedingly rare for a child younger than 1 year to have a malignant neoplasm as also seen in our series [6,7]. The presenting symptoms of these patients can significantly overlap with benign conditions. Complaints of an abdominal mass or precocious puberty are statistically more often associated with an underlying malignancy (Table 2) [1,2,6-10]. Evidence of unusual hormonal activity is more often associated with either estrogen production from a juvenile granulosa cell tumor or androgen production with resultant virilization from a Sertoli-Leydig tumor [9,11,12]. However, a similar level of hormone production cannot be infrequently seen with autonomously functioning benign follicular cysts [13]. β-HCG, αFP, and CA-125 are all potential tumor markers for ovarian malignancies. Germ cell tumors will often produce αFP and/or β-HCG, depending on their subgrouping
133 [14,15]. Dysgerminomas can result in elevations of serum lactic dehydrogenase, β-HCG, or CA-125 [12]. Yolk Sac and Sertoli-Leydig cell tumors are both associated with αFP elevations. CA-125 is most commonly associated with epithelial type tumors [10]. However, the caveat is that tumor markers do not exclude the possibility of malignancy when they are not elevated. Thus, tumor markers are not definitive measurements preoperatively but are of greatest value, if elevated, in the ability to monitor postoperatively for complete resection of disease, as well as to detect relapse [16,17]. Tumor markers should be drawn preoperatively whenever possible, but elevations can still be detected immediately postoperatively owing to the relatively long half-lives of both αFP and β-HCG [16]. Preoperative imaging will often reveal a pelvic mass, easily detected by ultrasound, CT scan, or magnetic resonance imaging, depending on the initial complaints prompting the radiologic evaluation. Malignant masses were generally thought to be greater than 5 cm [6,18], whereas others advocate the use of more than 8 cm to be the size cutoff for concern of malignancy [3,4]. In our population, all malignancies exceeded 6 cm, whereas 8 cm or more allowed for a statistically significant odds ratio with the greatest accuracy (Table 2). The caveat is that in our series, using the 8-cm cutoff, 2 (4.3%) of 46 malignancies would have been missed by size alone. Radiologic studies are not always able to distinguish benign lesions from malignancies. Classically, solid masses are thought to have a higher risk of malignancy, although heterogeneous and cystic masses can also harbor underlying malignant neoplasm [1]. In our series, a solid mass was not commonly encountered in our population. We noted that although solid masses, either by ultrasound or by CT scan, had the highest frequency (33%-60%) of malignancy, solid features were only evident in 14% to 15% of all malignancies. Conversely, most radiologic reads identified either a heterogeneous or cystic mass, which were associated with malignancy in only 15% to 20% and 4% to 5% of the time, respectively, which is consistent with previous reports [2]. Although findings of a cystic lesion may be relatively reassuring, heterogenous masses should still be approached with a heightened suspicion for malignancy, especially if additional concerning factors are present. In our series, germ cell tumors accounted for the majority (50%) of pediatric ovarian malignancies, which is consistent with the literature and in up to two thirds of those 9- to 14year-olds [7,12,14]. Sex cord stromal tumors make up a third of our population. The 17% of our series with epithelial tumors are also within previously reported ranges and consisted of the low malignant potential serous borderline tumors [10,12]. Our series, although limited in size, is the largest single institution review and has the ability to analyze features of benign ovarian masses compared with malignancies, which is not possible in national oncology databases. Although ovarian malignancies make up a small percentage of childhood malignancies, they still account for roughly 10%
134 of ovarian masses within the pediatric population, making the evaluation of an ovarian mass a diagnostic dilemma. Our findings suggestive of a potential malignancy include masses in those aged 1 to 8 years of age and a chief complaint of a mass or precocious puberty. Imaging findings of a mass greater than 8 cm, or a mass that is solid in character, are clearly concerning for malignancy. Laboratory investigations showing elevated β-HCG, αFP, or CA-125 are worrisome for malignant neoplasm but are not definitive if normal. Although at the time of surgery, a biopsy with frozen section can resolve some dilemmas, it does have the risk of upstaging a potential tumor, which is the rationale of not performing a frozen section unless faced with bilaterality. Ovarian preservation and salvage techniques could potentially be positively impacted by better risk stratification of ovarian masses preoperatively.
References [1] Cass D, Hawkings E, Brandt ML, et al. Surgery for ovarian masses in infants, children, and adolescents: 102 consecutive patients treated in a 15-year period. J Pediatr Surg 2001;36:693-9. [2] Brown MF, Hebra A, McGeehin K, et al. Ovarian masses in children: a review of 91 cases of malignant and benign masses. J Pediatr Surg 1993;28:930-2. [3] Hayes-Jordan A. Surgical management of the incidentally identified ovarian mass. Sem Pediatr Surg 2005;14:106-10. [4] Bristow RE, Nugent AC, Zahurak ML, et al. Impact of surgeon specialty on ovarian-conserving surgery in young females with an adnexal mass. J Adolesc Health 2006;39:411-6.
S.C. Oltmann et al. [5] Brookfield KF, Cheung MC, Koniaris LG, et al. A population-based analysis of 1,037 malignant ovarian tumors in the pediatric population. J Surg Res 2009 [in press]. [6] Billmire D, Vinocur C, Rescorla F, et al. Outcome and staging evaluation in malignant germ cell tumors of the ovary in children and adolescents: an intergroup study. J Pediatr Surg 2004;39:424-9. [7] Templeman C, Fallat ME, Blinchevsky A, et al. Noninflammatory ovarian masses in girls and young women. Obstet Gynecol 2000;96: 229-33. [8] Einarsson JI, Edwards CL, Zurawin RK. Immature ovarian teratoma in an adolescent: a case report and review of the literature. J Pediatr Adolesc Gynecol 2004;17:187-9. [9] Niedziela M. Virilizing ovarian tumor in a 14-year-old female with prior familial multinodular goiter. Pediatr Blood Cancer 2008;51: 543-5. [10] Morowitz M, Huff D, von Allmen D. Epithelial ovarian tumors in children: a retrospective analysis. J Pediatr Surg 2003;38:331-5. [11] Calaminus G, Wessalowski R, Harms D, et al. juvenile granulosa cell tumors of the ovary in children and adolescents: results from 33 patients registered in a prospective cooperative study. Gynecol Oncol 1997;65:447-52. [12] Von Allmen D. Malignant lesions of the ovary in childhood. Semin Pediatr Surg 2005;14:100-5. [13] Brandt ML, Helmrath MA. Ovarian cysts in infants and children. Semin Pediatr Surg 2005;14:100-5. [14] Baranzelli MC, Bouffet E, Quintana E, et al. Non-seminomatous ovarian germ cell tumours in children. Eur J Can 2000;36:376-83. [15] Schultz KAP, Sencer SF, Messinger Y, et al. Pediatric ovarian tumors: a review of 67 cases. Pediatr Blood Cancer 2005;44:167-73. [16] Childrens Oncology Group. Treatment of children with germ cell tumors handbook. www.childrensoncologygroup.org; 2004. [17] Pectasides D, Pestasides E, Kassanos D. Germ cell tumors of the ovary. Cancer Treat Rev 2008;34:427-41. [18] Tsai JY, Saigo PE, Brown C, et al. Diagnosis, pathology, staging, treatment, and outcome of epithelial ovarian neoplasia in patients age b 21 years. Cancer 2001;91:2065-70.
www.elsevier.com/locate/jpedsurg
Can we preoperatively risk stratify ovarian masses for malignancy? Sarah C. Oltmann a , Nilda Garcia a , Robert Barber b , Rong Huang b , Barry Hicks a , Anne Fischer a,⁎ a
University of Texas Southwestern Medical Center, Dallas, TX 75235, USA Children's Medical Center, Dallas, TX 75235, USA
b
Received 30 September 2009; accepted 6 October 2009
Key words: Ovarian mass; Ovarian malignancy; Pediatric surgery; Gynecology
Abstract Purpose: Given a 10% malignancy rate in pediatric ovarian masses, what preoperative factors are helpful in distinguishing those at higher risk to risk stratify accordingly? Methods: After institutional review board approval (IRB#022008-095), a 15½-year retrospective review of operative ovarian cases was performed. Results: A total of 424 patients were identified, with a mean age 12.5 years (range, 1 day to 19 years), without an age disparity between benign (12.54 years, 89%) and malignant (11.8 years, 11%) cases. The 1- to 8-year age group had the highest percentage of malignancies (22%; odds ratio [OR], 3.02; 95% confidence interval [CI], 1.33-6.86). A chief complaint of mass or precocious puberty versus one of pain had an OR for malignancy of 4.84 and 5.67, respectively (95% CI, 2.48-9.45 and 1.60-20.30). Imaging of benign neoplasms had a mean size of 8 cm (range, 0.9-36 cm) compared with malignancies at 17.3 cm (6.2-50 cm, P b .001). An ovarian mass size of 8 cm or longer on preoperative imaging had an OR of 19.0 for malignancy (95% CI, 4.42-81.69). Ultrasound or computed tomographic findings of a solid mass, although infrequent, were most commonly associated with malignancy (33%-60%), compared with reads of heterogeneous (15%-21%) or cystic (4%-5%) lesions. The malignancies (n = 46) included germ cell (50%, n = 23), stromal (28%, n = 13), epithelial (17%, n = 8), and other (4%, n = 2). Tumor markers obtained in 71% of malignancies were elevated in only 54%, whereas 6.5% of those sent in benign cases were similarly elevated. Elevated beta-human chorionic gonadotropin (β-HCG), alpha fetoprotein (αFP), and cancer antigen 125 (CA-125) were significantly associated with malignancy (P b .02) and an elevated carcinoembryonic antigen (CEA) was not (P = .1880). Conclusion: This reported series of pediatric ovarian masses demonstrates that preoperative indicators that best predict an ovarian malignancy are a complaint of a mass or precocious puberty, a mass exceeding 8 cm or a mass with solid imaging characteristics. Those patients aged 1 to 8 years have the greatest incidence of malignancy. Tumor markers, positive or negative, were not conclusive in all cases but useful for postoperative surveillance. © 2010 Elsevier Inc. All rights reserved.
Presented at the 40th Annual Meeting of the American Pediatric Surgical Association, Fajardo, Puerto Rico, May 28-June 1, 2009. ⁎ Corresponding author. Tel.: +1 214 456 6040; fax: +1 214 456 6320. E-mail address: [email protected] (A. Fischer). 0022-3468/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.jpedsurg.2009.10.022
Although ovarian malignancy is relatively rare in the pediatric population constituting only 1% of all childhood malignancies, ovarian masses in childhood do have a 10%
Stratified ovarian masses and malignancy malignancy rate [1,2]. Strategies for ovarian salvage have been slow to be implemented given the unclear malignancy risk of a given ovarian mass upon presentation; high oophorectomy rates for nonmalignancies have continued to be reported despite the endorsement of ovarian salvage in the recent literature [1,3]. It is unclear how best to determine preoperatively which subset of patients with an ovarian lesion is at higher risk for malignancy, particularly because these factors may impact the decision to perform an oophorectomy at the time of surgery. Significant overlap in the clinical presentation exists between benign and malignant pathology, and dogma surrounding the risk of malignancy in ovarian masses has little evidence-based support within the current literature. In an effort to better ascertain preoperative indicators of malignancy compared with benign masses, we analyzed our experience with pediatric ovarian pathology at a major academic center to see if patient characteristics, laboratory values, and radiologic parameters exist, which can identify those patients most at risk for ovarian malignancy.
1. Methods After institutional review board approval (IRB#-022008095), International Classification of Diseases-9th revision (ICD-9) codes and pathology database were used to identify those patients from birth to 19 years of age surgically managed for ovarian findings at a major academic center consisting of a large, free-standing children's hospital and a major metropolitan county hospital from January 1, 1993 to June 30, 2008. Those patients with tubo-ovarian abscesses or isolated ectopic pregnancy were excluded. A retrospective review of admission notes, radiology reports, operative reports, laboratory results, and pathology reports was performed by a single author for internal consistency. Descriptive analyses were performed using frequency distributions, means, and SEM. χ2 and Fisher's Exact tests were used to compare proportions between 2 groups. MantelHaenszel method was used to calculate relative risk and its confidence intervals. Two-sample t test was used to compare means between the 2 groups. The significance level was set as P value less than .05. GraphPad Prism 5.01 and SAS 9.1 were used for statistical calculation.
131 Table 1 Demographics of malignant versus benign pediatric ovarian masses
No. of Patients No. of Cases Age (y), mean ± SEM Chief complaint Mass Pain Precocious puberty Prenatal imaging Incidental finding Per Orem (PO) intolerance Other/not documented WBC N12,000 Tumor markers percent abnormal findings (abnormal/drawn) beta-human chorionic gonadotriopin (β-HCG) a alpha fetoprotein (αFP) cancer antigen 125 (CA-125), carcinoembryonic antigen (CEA) Mean size by imaging (cm)
Malignant
Benign
46 (11%) 48 (11%) 11.8 ± 0.64
378 (89%) 390 (89%) 12.54 ± 0.26
22 (46%) 20 (42%) 4 (8%) 0 0 0
58 (15%) 255 (65%) 9 (2%) 11 (3%) 35 (9%) 3 (1%)
2 (4%)
19 (5%)
7 (15%) 54% (19/35)
59 (15%) 6.5% (8/122)
.92 b.001
22% (6/27)
0% (0/107)
b.001
52% (13/25) 71% (5/7) 25% (2/8) 17.3 (6.2-50)
1.9% (1/54) 22% (7/32) 5% (1/20) 8 (0.9-36)
b.001
.64 b.001 .001 .02
.83
b.02 .188 b.001
WBC indicates white blood cell. a Those patients who were pregnant at the time of evaluation were excluded from this subset.
P b .01; Fig. 1) in comparison to 10% seen in older girls. This finding was further substantiated by a 3-fold greater odds ratio for malignancy in the 1- to 8-year group compared with the 15- to 19-year olds (Table 2).
2. Results A total of 424 patients were identified, who underwent a total of 438 operative cases with ovarian findings. Forty-six patients (11%) were found to have malignant disease, with 2 requiring reoperation for local recurrence. The mean (SEM) age of those patients with malignancy was 11.8 (±0.64) years, similar to the benign population (12.54 ± 0.26 years; Table 1). However, when looking at the relative percentage of malignancy by age intervals, the subset aged 1 to 8 years had the highest percentage of malignant neoplasms (22%,
P
Fig. 1
Pathology distribution by age group.
132 Table 2
S.C. Oltmann et al. Table 3
Odds ratios for malignancy Odds ratio
Chief complaint Mass 4.84 Precocious puberty 5.67 Pain 1.0 Age at presentation (y) b1 0.364 1-8 3.02 9-14 1.338 15-19 1.0 Imaging characteristic (Ultrasound) Heterogeneous 3.62 Cystic 1.0 Solid 10.13 Imaging characteristic (CT) Heterogeneous 6.84 Cystic 1.0 Solid 39.0 Mass size by imaging (cm) ≥5 18.9 ≥6 31.4 ≥8 19.0
Malignancy by cell type
95% confidence interval 2.48-9.45 1.60-20.03
0.046-2.853 1.33-6.86 0.664-2.695
1.15-11.36 1.83-56.05 1.49-31.36
n Germ cell Yolk sac Immature teratoma Mixed germ cell Dysgerminoma Sex cord stromal Juvenile granulosa cell Sertoli-Leydig Indeterminant sex cord Epithelial Borderline serous tumor Carcinoma Other Lymphoma Leukemia
23 (50%) 4 5 9 5 13 (28%) 8 4 1 8 (17%) 6 2 2 (4%) 1 1
3.99-381 1.14-314.6 1.90-520.27 4.42-81.69
The presenting symptoms for those patients with malignancy were equally divided between a mass (n = 22, 46%) and pain (n = 20, 42%), whereas precocious puberty (n = 4, 8%) accounted for the rest (Table 1). However, because pain was also the primary complaint in the majority (65%) of the benign population, pain as a symptom was not a distinguishing characteristic. In contrast, those patients with either a chief complaint of mass or precocious puberty were significantly more likely to have a malignancy than those patients with a complaint of pain (Table 2). On laboratory evaluation, a leukocytosis exceeding 12,000 was found with equal incidence in both benign and malignant cases (Table 1). Although elevated tumor markers (β-HCG, αFP, CA-125) were highly associated with malignancy (P b .02), they were elevated only 54% of the time. Thus, in a comparable number of incidences (46%), the tumor markers were actually normal in the setting of an ovarian malignancy and surprisingly elevated in 6.5% of benign lesions. Elevated CEA, commonly seen in malignancies in adults, was less commonly assessed or detected in any of the malignancies (in 2 of 8 cases) and was not found to be associated with malignancy (P = .188). Patients were most commonly imaged with ultrasound and/or computed tomographic (CT) scan. Thirty patients had ultrasound preoperatively, 33 patients had CT scan, and 4 underwent magnetic resonance imaging evaluation. Malignant lesions all exceeded 6 cm, with an average size of 17.3 cm (range, 6.2-50 cm), significantly larger than the benign lesions averaging 8 cm (range, 0.9-36 cm; P b .001). We statistically analyzed 3 different diameters to determine which would be the most helpful clinically: 5 cm, the traditional cutoff; 6 cm, the size that all
malignancies exceeded in our series; and 8 cm [3,4]. Although the odds ratio is greatest for those diameters of 6 cm or wider at 31.4, the wide 95% confidence interval of 1.90 to 520.27 is less precise. Thus, the ≥ 8-cm-or-more cutoff was not only significant with an odds ratio of 19.0, but also had a narrower and thus more accurate confidence interval (Table 2). In addition to evaluating the mean lesion size by imaging, the nature of the lesion or mass as described by the radiologist is also important. Forty-three percent of solid lesions (6 of 14) imaged by ultrasound or CT scan were associated with malignancy, whereas complex/heterogeneous and cystic masses were each associated with malignancy in only 18% (30 of 171) and 4% (6 of 139) of cases, respectively. Resultant odds ratios for malignancy in ovarian masses that were solid or heterogeneous masses by CT were 39 and 6.48, respectively, in comparison to a cystic lesion (Table 2). Likewise, similar odds ratios were seen upon analysis of ultrasound characteristics. Because treatment and prognosis can vary significantly by malignant cell type, the pathology was also noted (Table 3).
Fig. 2
Distribution of malignancy cell type by age group.
Stratified ovarian masses and malignancy Germ cell tumors comprised 50% of all the pediatric ovarian malignancies; sex cord stromal, 28%; epithelial, 17%; and other, 4%. This distribution changed with patient age (Fig. 2). Germ cell tumors accounted for most ovarian malignancies in all age groups and the greatest in those 9 to 14 years. Sex cord stromal tumors comprised 25% to 33% of malignancies in all groups over the age of 1. Epithelial tumors only occurred in those patients over the age of 12 years and presented more often as the less aggressive serous borderline tumor than the typical epithelial carcinoma seen in adults. One patient was found to have a recurrence of acute lymphoblastic leukemia in the ovary, and another patient had primary lymphoma of the ovary.
3. Discussion Ovarian malignancies in the pediatric population are rare. In a population-based study, age-adjusted incidence for malignant pediatric ovarian tumors varied 10-fold for each age interval from only 0.102 per 100,000/y in those girls 9 years or younger, 1.072 per 100,000/y in those age 10 to 19 years, to 11.446 per 100,000/y in the adult female population [5]. Although those patients 10 years and older may have a higher age-adjusted incidence in the populationbased study, this does not take into consideration the increased incidence of other benign ovarian lesions seen within this age group. Our study, which determined percentage of malignancies in the context of an ovarian mass, found the greatest percentage of malignancies in those patients aged 1 to 8 years. As demonstrated with our pathology distribution by age in Fig. 1, perimenarchal patients tend to have a greater number of benign lesions by way of cysts and neoplasms, whereas those patients hormonally quiescent are less likely to present with an ovarian mass at all. But when they do, malignancy may be present up to a quarter of the time. As a result, a higher index of suspicion for malignancy is needed in the patient aged 1 to 8 years, as indicated by a 3-fold increase in the odds ratio. Conversely, it is exceedingly rare for a child younger than 1 year to have a malignant neoplasm as also seen in our series [6,7]. The presenting symptoms of these patients can significantly overlap with benign conditions. Complaints of an abdominal mass or precocious puberty are statistically more often associated with an underlying malignancy (Table 2) [1,2,6-10]. Evidence of unusual hormonal activity is more often associated with either estrogen production from a juvenile granulosa cell tumor or androgen production with resultant virilization from a Sertoli-Leydig tumor [9,11,12]. However, a similar level of hormone production cannot be infrequently seen with autonomously functioning benign follicular cysts [13]. β-HCG, αFP, and CA-125 are all potential tumor markers for ovarian malignancies. Germ cell tumors will often produce αFP and/or β-HCG, depending on their subgrouping
133 [14,15]. Dysgerminomas can result in elevations of serum lactic dehydrogenase, β-HCG, or CA-125 [12]. Yolk Sac and Sertoli-Leydig cell tumors are both associated with αFP elevations. CA-125 is most commonly associated with epithelial type tumors [10]. However, the caveat is that tumor markers do not exclude the possibility of malignancy when they are not elevated. Thus, tumor markers are not definitive measurements preoperatively but are of greatest value, if elevated, in the ability to monitor postoperatively for complete resection of disease, as well as to detect relapse [16,17]. Tumor markers should be drawn preoperatively whenever possible, but elevations can still be detected immediately postoperatively owing to the relatively long half-lives of both αFP and β-HCG [16]. Preoperative imaging will often reveal a pelvic mass, easily detected by ultrasound, CT scan, or magnetic resonance imaging, depending on the initial complaints prompting the radiologic evaluation. Malignant masses were generally thought to be greater than 5 cm [6,18], whereas others advocate the use of more than 8 cm to be the size cutoff for concern of malignancy [3,4]. In our population, all malignancies exceeded 6 cm, whereas 8 cm or more allowed for a statistically significant odds ratio with the greatest accuracy (Table 2). The caveat is that in our series, using the 8-cm cutoff, 2 (4.3%) of 46 malignancies would have been missed by size alone. Radiologic studies are not always able to distinguish benign lesions from malignancies. Classically, solid masses are thought to have a higher risk of malignancy, although heterogeneous and cystic masses can also harbor underlying malignant neoplasm [1]. In our series, a solid mass was not commonly encountered in our population. We noted that although solid masses, either by ultrasound or by CT scan, had the highest frequency (33%-60%) of malignancy, solid features were only evident in 14% to 15% of all malignancies. Conversely, most radiologic reads identified either a heterogeneous or cystic mass, which were associated with malignancy in only 15% to 20% and 4% to 5% of the time, respectively, which is consistent with previous reports [2]. Although findings of a cystic lesion may be relatively reassuring, heterogenous masses should still be approached with a heightened suspicion for malignancy, especially if additional concerning factors are present. In our series, germ cell tumors accounted for the majority (50%) of pediatric ovarian malignancies, which is consistent with the literature and in up to two thirds of those 9- to 14year-olds [7,12,14]. Sex cord stromal tumors make up a third of our population. The 17% of our series with epithelial tumors are also within previously reported ranges and consisted of the low malignant potential serous borderline tumors [10,12]. Our series, although limited in size, is the largest single institution review and has the ability to analyze features of benign ovarian masses compared with malignancies, which is not possible in national oncology databases. Although ovarian malignancies make up a small percentage of childhood malignancies, they still account for roughly 10%
134 of ovarian masses within the pediatric population, making the evaluation of an ovarian mass a diagnostic dilemma. Our findings suggestive of a potential malignancy include masses in those aged 1 to 8 years of age and a chief complaint of a mass or precocious puberty. Imaging findings of a mass greater than 8 cm, or a mass that is solid in character, are clearly concerning for malignancy. Laboratory investigations showing elevated β-HCG, αFP, or CA-125 are worrisome for malignant neoplasm but are not definitive if normal. Although at the time of surgery, a biopsy with frozen section can resolve some dilemmas, it does have the risk of upstaging a potential tumor, which is the rationale of not performing a frozen section unless faced with bilaterality. Ovarian preservation and salvage techniques could potentially be positively impacted by better risk stratification of ovarian masses preoperatively.
References [1] Cass D, Hawkings E, Brandt ML, et al. Surgery for ovarian masses in infants, children, and adolescents: 102 consecutive patients treated in a 15-year period. J Pediatr Surg 2001;36:693-9. [2] Brown MF, Hebra A, McGeehin K, et al. Ovarian masses in children: a review of 91 cases of malignant and benign masses. J Pediatr Surg 1993;28:930-2. [3] Hayes-Jordan A. Surgical management of the incidentally identified ovarian mass. Sem Pediatr Surg 2005;14:106-10. [4] Bristow RE, Nugent AC, Zahurak ML, et al. Impact of surgeon specialty on ovarian-conserving surgery in young females with an adnexal mass. J Adolesc Health 2006;39:411-6.
S.C. Oltmann et al. [5] Brookfield KF, Cheung MC, Koniaris LG, et al. A population-based analysis of 1,037 malignant ovarian tumors in the pediatric population. J Surg Res 2009 [in press]. [6] Billmire D, Vinocur C, Rescorla F, et al. Outcome and staging evaluation in malignant germ cell tumors of the ovary in children and adolescents: an intergroup study. J Pediatr Surg 2004;39:424-9. [7] Templeman C, Fallat ME, Blinchevsky A, et al. Noninflammatory ovarian masses in girls and young women. Obstet Gynecol 2000;96: 229-33. [8] Einarsson JI, Edwards CL, Zurawin RK. Immature ovarian teratoma in an adolescent: a case report and review of the literature. J Pediatr Adolesc Gynecol 2004;17:187-9. [9] Niedziela M. Virilizing ovarian tumor in a 14-year-old female with prior familial multinodular goiter. Pediatr Blood Cancer 2008;51: 543-5. [10] Morowitz M, Huff D, von Allmen D. Epithelial ovarian tumors in children: a retrospective analysis. J Pediatr Surg 2003;38:331-5. [11] Calaminus G, Wessalowski R, Harms D, et al. juvenile granulosa cell tumors of the ovary in children and adolescents: results from 33 patients registered in a prospective cooperative study. Gynecol Oncol 1997;65:447-52. [12] Von Allmen D. Malignant lesions of the ovary in childhood. Semin Pediatr Surg 2005;14:100-5. [13] Brandt ML, Helmrath MA. Ovarian cysts in infants and children. Semin Pediatr Surg 2005;14:100-5. [14] Baranzelli MC, Bouffet E, Quintana E, et al. Non-seminomatous ovarian germ cell tumours in children. Eur J Can 2000;36:376-83. [15] Schultz KAP, Sencer SF, Messinger Y, et al. Pediatric ovarian tumors: a review of 67 cases. Pediatr Blood Cancer 2005;44:167-73. [16] Childrens Oncology Group. Treatment of children with germ cell tumors handbook. www.childrensoncologygroup.org; 2004. [17] Pectasides D, Pestasides E, Kassanos D. Germ cell tumors of the ovary. Cancer Treat Rev 2008;34:427-41. [18] Tsai JY, Saigo PE, Brown C, et al. Diagnosis, pathology, staging, treatment, and outcome of epithelial ovarian neoplasia in patients age b 21 years. Cancer 2001;91:2065-70.