Is Risk Malignancy Index a Useful Tool for Predicting Malignant Ovarian Masses in Developing Countries?

Introduction. Risk of Malignancy Index (RMI) is widely studied for prediction of malignant pelvic masses in Western population. However, little is known regarding its implication in the developing countries. The objective of this study is to determine how accurately the RMI can predict the malignant pelvic masses. Materials and Methods. The study is a retrospective review of patients attending the gynecological clinic between January 2004 and December 2008 with adnexal masses. Information on demographic characteristics, ultrasound findings, menopausal status, CA125, and histopathology was collected. RMI score for each patient in the study group was calculated. Results. The study group included a total of 283 patients. Analysis of the individual parameters of RMI revealed that ultrasound was the best predictor of malignancy with a sensitivity, specificity, and positive likelihood ratio of 78.3%, 81.5%, and 4.2, respectively. At a standard cut-off value of 250, RMI had a positive likelihood ratio of 8.1, while it was 6.8 at a cut-off of 200, albeit with comparable sensitivity and specificity. Conclusion. RMI is a sensitive tool in predicting malignant adnexal masses. A cut-off of 200 may be suitable in developing countries for triaging and early referral to tertiary care centers.


Introduction
Ovarian masses are a frequent cause of gynecological consults and are often detected during imaging studies or exploratory surgery for evaluation of abdominal or pelvic pain syndromes. They occur across age groups and could result from benign or malignant disease. With more than 250,000 new cases reported every year, ovarian malignancies represent the fourth commonest cause of cancer deaths worldwide [1]. They also have the lowest 5-year survival rate (30-50%) among all gynecological cancers [2]. A recent report indicated an increasing incidence of ovarian cancers in the developing world, compared to the developed countries [3].
Early identification of ovarian carcinomas and referral to a gyneco-oncologist can facilitate accurate staging of the disease and optimal cytoreductive treatment, enhancing patient survival [4,5]. Histopathology remains the diagnostic gold standard for this cancer, and a definitive biomarker has not been identified yet. Risk of Malignancy Index (RMI), which considers the serum CA125 level, menopausal status, and ultrasonographic findings in predicting malignant pelvic masses, is widely employed in the developed countries [6]. However, its utility in risk prediction in the developing countries is currently unknown.
The present study evaluated how accurately the RMI can predict the risk of malignant pelvic masses, among patients with an ovarian mass.

Material and Methods
After the approval of our institutional review board we conducted a retrospective review of the case files of patients with adnexal masses who attended the Gynecological Clinic at the Aga Khan University, Karachi, Pakistan, between January 2004 and December 2008. The International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) criteria were used to identify adnexal masses. Patients with advanced disease were excluded from the study. We collected information on demographic characteristics, ultrasonographic findings, menopausal status, serum CA125 level, and histopathology. The RMI for each patient was calculated using the standard formula [6]. 3.2.6. Ultrasound Score. We assigned scores of 0 (absence of specific findings), 1 (presence of one finding), or 3 (two or more findings) to the subjects, depending on the ultrasound findings. One hundred and nineteen (42.3%) cases had an ultrasound score of 1, while lesions of 88 (31.3%) and 74 (26.3%) patients were scored 0 and 3, respectively. Of the 207 (73.6%) patients with an ultrasound score 1, 196 (69.7%) had benign disease, while 8 (2.8%) and 3 (1%) had malignant and borderline disease, respectively. Seventy-four (26.3%) patients in our series had an ultrasound score of 3, and among them, 41 (14.5%) had benign, 29 (10.3%) had malignant, and 4 (1.4%) had borderline disease, respectively.

Risk Stratification Based on RMI Scores.
We assessed the distribution of benign, borderline, and malignant ovarian cancers when the patients were categorized based on their RMI scores. In order to identify the RMI score that was an effective risk predictor, we studied the sensitivity and specificity of RMI scores at four levels, namely, ≤100, ≤150, ≤200, and ≤250. The sensitivity, specificity, and positive and negative predictive values of RMI score at each of these levels are summarized in Table 3.
To find out the RMI score that could most effectively classify the disease, we calculated the sensitivity, specificity, positive predictive value, negative predictive value, and the likelihood ratios at RMI cut-off levels of 100, 150, 200, and 250. A comparison of the diagnostic indices with these cutoffs is shown in Table 3.
As shown in Table 3, an RMI of 250 yielded the ideal combination of sensitivity (54.05), specificity (93.4), positive predictive value (55.5), negative predictive value (93.06), and positive (8.1) and negative (0.49) likelihood ratios. Though cut-offs of 100 and 150 showed higher sensitivity in detecting malignant disease, they had lower specificity, positive predictive value, and likelihood ratios, compared to 250.
We also compared the diagnostic performance of RMI scores >250 against CA125 levels >35 U/mL, ultrasound score of 3 and menopausal score of 3. Table 4 summarizes the findings from this analysis. Among the three criteria, an ultrasound score of 3 had the highest sensitivity (78.3%), while an RMI score ≥250 had the highest specificity (93.4%). The latter also had the highest positive predictive value of 55.5%, while negative predictive value was highest for an ultrasound score of 3 (96.1%). The positive likelihood ratio was highest for RMI score ≥250, while a score of 100 had the least negative likelihood ratio (0.39).

Discussion
About 10% of women undergo exploratory surgery for evaluation of ovarian masses during their lifetime [7]. Prompt identification of ovarian malignancies and referral to a gyneco-oncologist can enhance the patient survival rates [8], but a single method which can accurately predict ovarian malignancy is still unavailable. Herein we report that the multiparametric RMI score can be a useful tool in prediction of malignant ovarian disease, in low-resource settings.
The mean age of the patients with ovarian mass in our study was 36.87 years (range, 8 to 85 years). This is slightly higher than that reported in a similar study by Akdeniz et al. in 2009 [9].
In our study, 13.2% of the patients with an ovarian mass had malignant disease. Thirty-five percent of malignancies occurred in postmenopausal patients and 7.9% among the premenopausal patients. The data seem to agree with earlier reports of similar incidence rates and preponderance in postmenopausal patients [9][10][11][12].
Ultrasonography is widely appreciated as the best imaging method for evaluation of ovarian pathology. Several groups have reported higher sensitivity, specificity, and positive predictive values for this method (Agarwal et al., 2011, and references therein). In our study, an ultrasound score of 3 had the highest sensitivity (78.3%) and negative predictive value (96.1%) and the least negative likelihood ratio (0.26), among the parameters evaluated.
Several candidate biomarkers and their combinations have been employed in assessing the risk of ovarian malignancies, albeit with varying efficiency [13]. Serum CA125 level is widely appreciated as a useful biomarker for estimating the risk of ovarian cancer, though other gynecological pathology can also increase its levels. Myers et al. [14] have earlier reported sensitivity and specificity of less than 80%, for this marker, in the prediction of ovarian cancers. Simsek et al. (2014) [15] reported a sensitivity of 78.6% and specificity of 63.5% for a CA125 cut-off of 35 U/mL. Another report indicated a sensitivity of 88% and specificity of 97% for CA125 at a higher cut-off of 88 U/mL [12]. In our study, CA125 levels ≥35 U/mL had a sensitivity of 70.2%, specificity of 67.6%, positive predictive value of 24.7%, negative predictive value of 93.7, and positive and negative likelihood ratios of 2.1 and 0.44, respectively. We suggest that a higher prevalence of inflammatory and nonspecific uterine and ovarian pathology might have contributed to elevated CA125 levels in the majority of our patients and thus its low diagnostic performance in the detection of malignant ovarian disease.
Rao (2014) [16] has recently reported higher sensitivity, specificity, and positive and negative predictive values for a postmenopausal score of 3. In our study, this parameter had a higher specificity and negative predictive value, but lower sensitivity and positive predictive values in assessing malignancy risk.
RMI was first proposed by Jacobs et al. and is calculated from the serum CA125 antigen level, menopausal status, and ultrasonographic findings [6]. Several retrospective and prospective studies have reported it to be the best available tool for triage and referral of ovarian malignancies [17,18]. Its utility as a diagnostic tool depends on the prevalence of malignancy in the study population [15]. We observed a low prevalence of malignancy (13.2%) among our study group, significantly lesser than some of the earlier reports of 30-43% [6,17,19]. Jacobs et al. (1990) [6], studying 143 patients, reported a sensitivity of 85.4% and specificity of 96.9% for this method, with a cut-off of 200. Subsequently, several groups have reported its superior sensitivity and specificity in estimating the risk of ovarian malignancy, compared to other parameters [19][20][21][22][23][24][25]. The RMI cut-offs in many studies ranged from 25 to 250 (reviewed in Geomini et al., 2009) [18]. Most studies reported an increased diagnostic accuracy and performance with an RMI cut-off of 200 [6,16,19,20,22,24,[26][27][28][29][30][31][32]. A recent study reported a sensitivity of 89.5%, specificity of 96.2%, positive predictive value of 77.3%, and negative predictive value of 98.4% [11], when a higher RMI cut-off of 238 was used for the screening. Yamamoto et al. (2009) [25] reported a sensitivity and specificity of 75% and 91%, respectively, using a cut-off of 450. The best performance in the present study was seen with an RMI cut-off of 250, and the low sensitivity (54.5%) and high specificity (93.4%) observed were comparable to the majority of earlier reports that employed a similar cut-off [6,19,20,22,26,[29][30][31][32][33][34][35].
We conclude that, in the absence of a definitive biomarker, the multiparametric Risk of Malignancy Index serves as a very useful tool for identification of malignant ovarian disease and their prompt triage and referral to expert care.