Systematic review and meta-analysis of imaging differential diagnosis of benign and malignant ovarian tumors

Introduction

The incidence of ovarian cancer is very high, and it is also one of the main causes of female deaths due to tumors in the reproductive system. Current reports show that ovarian cancer is the cause of death of more than 180,000 women worldwide every year (1). High heterogeneity is a characteristic of ovarian cancer, which can be divided into epithelial tumors, germ cell tumors, and sex cord-stromal tumors according to pathology (2). Among them, epithelial ovarian cancer is the most common, accounting for about 90% of all cases (3). Common risk factors for the onset of ovarian cancer include family genetic history, exercise, lifestyle, diet, fertility and breastfeeding, menstruation, body mass index, gynecological related diseases, hormone replacement therapy, and even psychological factors (4-8).

Ovarian cancer is a highly aggressive ovarian tumor. Because ovarian cancer rarely observes specific clinical manifestations or signs in the early stage, it is easy to be overlooked in the early stage and cannot be diagnosed. Studies have shown that only 20–25% of patients can be detected and diagnosed in the early stages of ovarian cancer (9). Additionally, approximately 60% of patients with ovarian cancer are unfortunate, and they are diagnosed at an advanced stage (10). Patients diagnosed at an advanced stage usually have a poor prognosis, and the 5-year survival rate is often <30% (10). Benign tumors often lack typical symptoms, such as occasional bloating and abdominal masses in the lower abdomen. Malignant tumors grow rapidly and have irregular masses. Systemic symptoms, such as fever, weakness, loss of appetite, and weight loss, may appear in a short period. Benign lesions can be treated with surgery to achieve a good prognosis. Malignant lesions can be treated with multiple surgical and chemotherapy programs to control the condition, but the prognosis is usually poor when ovarian cancer is in the middle and late stages.

Clinical symptoms combined with blood tumor marker tests, ultrasound examinations, computed tomography (CT) scans, magnetic resonance imaging (MRI) scans, etc., can often make the diagnosis of ovarian cancer, but the differentiation of benign and malignant requires further judgment. Comparative studies of the sensitivity and specificity of routine examinations are extremely important for the differentiation between benign and malignant ovarian tumors. Clinical consideration needs to be directed to the diagnosis of the disease, and the formulation and implementation of treatment plans. Spiral CT and MRI are 2 commonly used diagnostic methods, which are easy to operate and have strong feasibility. The main purpose of this study was to analyze the value of CT and MRI in the differential diagnosis of benign and malignant ovarian tumors.

We present the following article in accordance with the PRISMA reporting checklist (https://gs.amegroups.com/article/view/10.21037/gs-21-889/rc).

Methods

Search strategy

We searched for articles that reported on using MRI or CT to distinguish between benign and malignant ovarian tumors, and then included these articles in the study. The following English biomedical databases were searched: PubMed, Cochrane Central Register of Controlled Trials, Embase, and Web of Science. The following major Chinese biomedical databases were searched: China National Knowledge Infrastructure, Wanfang, and Weipu. The following search terms were used: “CT” or “Computed Tomography” or “MRI” or “Magnetic Resonance imaging” and “ovarian cancer” or “ovarian tumor” or “ovarian neoplasm” or “adnexal mass” or “adnexal lesion.” The searches were limited to articles published in English or Chinese from January 2000 to September 2021.

Inclusion and exclusion criteria

The included studies should meet the following criteria: (I) be published between January 2000 and September 2021; (II) be a study that evaluated the accuracy of CT or/and MRI in distinguishing between benign and malignant ovarian tumors; (III) include data on the histopathological findings for the diagnosis of benign or malignant ovarian tumors; (IV) include research data on the sensitivity, specificity, and accuracy of CT or/and MRI; and (V) include data that enabled the true positive, false positive, false negative, and true negative values to be derived. Studies were excluded from the meta-analysis if they met any of the following exclusion criteria: (I) the sample in the study was <10 patients; (II) the magnetic field strength for MRI research was <1.5 T or not recorded; (III) other radiotracers were used for the CT research; (IV) the article concerned a study on which repeated articles had been published (in which case the latest published article was selected); and/or (V) the article reported on animal experiments.

Paper screening and data extraction

According to the pre-established inclusion criteria, 2 of the authors independently reviewed the titles and abstracts of all the retrieved articles. If the reviewers’ opinions differed, a third reviewer participated in the decision. The 2 reviewers independently extracted data from the selected articles after the screening. The extracted data included the first author, the country of the institute, the year of publication, and the CT or MRI diagnosis of the tumor (i.e., benign or malignant). Disagreements in the data extraction process were resolved via discussion until a consensus was reached. If the required information was unclear and the full text was not available, and the review author could not be contacted to obtain the relevant information, the article was excluded.

Quality assessment

The included articles were evaluated for quality, and the scope of the evaluation included patient selection, index testing, reference standards, flow, and time. Each indicator was evaluated based on the risk of bias, and the first 3 indicators were evaluated based on applicability issues. Each methodological quality was rated as “low risk”, “high risk”, or “unclear”. The Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) recommended by the Cochrane Collaboration was used to assess the methodological quality of the research.

Statistical analysis

The main outcome indicators extracted from the enrolled studies were calculated. A hierarchical logistic regression model and a summary receiver operating characteristic (SROC) model with 95% confidence and prediction area were used in this study to calculate the sensitivity and specificity of the included research indicators. The Chi-square test and Higgins I² test were used to assess the heterogeneity of the included studies. And a sensitivity analysis was carried out according to the Cochrane systematic review method for included studies. The positive likelihood ratio (PLR), negative likelihood ratio (NLR), and diagnostic odds ratio (DOR) were calculated using the bivariate generalized linear mixed model and the random-effects model. All analyses were performed using Stata 15.0 (StataCorp LLC, College Station, TX, USA), and set P<0.05 as the difference is statistically significant.

Results

Search results and study characteristics

A total of 537 records were confirmed in all databases, and 57 unusable records were eliminated. After screening, 391 records were used for retrieval, and finally 217 articles were retrieved; 126 articles in English and 91 articles in Chinese. After reading the full text and determining that 207 articles did not meet the research criteria, as the data were incomplete, or the article was a review or case report, 10 available articles remained. The specific process for the study is shown in Figure 1. In relation to the 10 remaining articles, the publication times ranged from 2008 to 2020. Basic data, such as the first author, country, publication year, journal, and identification auxiliary inspection method of the article, were extracted. The basic characteristics of the articles are shown in Table 1.

Figure 1 Flow diagram of the search, screening, and inclusion process.

Risk of bias and applicability judgments

In relation to the risk of bias assessment, the risk of bias for the patient selection in 2 articles was unclear, and the risk of bias for the patient selection in the other 8 articles was low. In relation to the index test bias, 3 articles had a high risk, 1 article had an unclear risk, and 6 articles had a low risk. The reference standard bias, flow, and time bias risks of all the articles were low. In the applicability bias assessment, the 3 indicators of patient selection bias, index test bias, and reference standard bias were all low risk. The results of the risk of bias assessment are shown in Figure 2.

Figure 2 Literature quality evaluation details.

Meta-analysis results

Overall analyses of CT

The meta-regression analysis showed that the pooled sensitivity of CT in distinguishing between benign and malignant ovarian tumors was 0.79 (95% CI: 0.70–0.87; Figure 3), and the pooled specificity was 0.87 (95% CI: 0.80–0.92; Figure 4). In addition, the sensitivity and specificity I² were 58.2% and 57.7%, respectively. The SROC area under the curve (SAUC) showed high accuracy (SAUC =0.9016; Figure 5). The closer the AUC of the SROC to 1.0, the more accurate CT was in distinguishing between benign and malignant ovarian tumors.

Figure 3 Sensitivity of studies: forest plot of sensitivities of 4 studies. Statistical method: inverse variance of the random-effects model. CI, confidence interval.

Figure 4 Specificity of studies: forest plot of specificities of 4 studies. Statistical method: inverse variance of the random-effects model. CI, confidence interval.

Figure 5 SROC curve for individual studies on the CT differential diagnosis of benign and malignant ovarian tumors. SROC, summary receiving operation characteristic; AUC, area under the curve; SE, standard error; CT: computed tomography.

The positive likelihood ratio is the multiple of the correctly diagnosed disease and the wrongly diagnosed disease in the diagnostic auxiliary examination. Thus, in this study, the greater the positive likelihood ratio, the higher the accuracy of CT in distinguishing between benign and malignant ovarian tumors. The analysis results show that the pooled positive likelihood ratio of CT in distinguishing between benign and malignant ovarian tumors was 5.26 (95% CI: 2.78–9.93), indicating that CT did not have a high value in distinguishing between benign and malignant ovarian tumors (Figure 6).

Figure 6 Forest plot of positive LR. Comparison of positive LR between the benign group and the malignant group. Statistical method: inverse variance of the random-effects model. LR, Likelihood ratio; CI, confidence interval.

The negative likelihood ratio is the multiple of the probability of incorrectly diagnosing a negative disease to the multiple of the negative probability of correctly diagnosing the disease. Thus, in this study, the smaller the negative likelihood ratio, the higher the accuracy of CT as an auxiliary examination in distinguishing and diagnosing benign and malignant ovarian tumors. The negative likelihood ratio summarized in this study was 0.26 (95% CI: 0.13–0.50), indicating that CT is not accurate in distinguishing between benign and malignant ovarian tumors (Figure 7).

Figure 7 Forest plot of negative LR. Comparison of negative LR between the benign group and the malignant group. Statistical method: inverse variance of the random-effects model. LR, Likelihood ratio; CI, confidence interval.

The diagnostic odds ratio reflects the closeness of the relationship between the diagnostic test results and the corresponding disease. The greater the diagnosis rate, the greater the ability to represent whether the disease is diagnosed by the diagnostic test or auxiliary examination. The results of this study showed that the combined diagnostic odds ratio of CT in distinguishing between benign and malignant ovarian tumors was 22.19 (95% CI: 7.54–65.30). Thus, the accuracy of CT in distinguishing between benign and malignant ovarian tumors cannot be considered sufficiently strong (Figure 8).

Figure 8 Forest plot of diagnostic odds ratio. Comparison of diagnostic odds ratio between the benign group and the malignant group. Statistical method: inverse variance of the random-effects model. OR, odds ratio; CI, confidence interval.

Overall analyses of MRI

The meta-regression analysis showed that the pooled sensitivity of MRI in distinguishing between benign and malignant ovarian tumors was 0.94 (95% CI: 0.91–0.95) (Figure 9), and the pooled specificity was 0.91 (95% CI: 0.90–0.93) (Figure 10). In addition, the sensitivity and specificity I² were 55.3% and 80.4%, respectively. The SROC of MRI showed higher accuracy (SAUC =0.9764) than CT (Figure 11). The closer the AUC value of the SROC to 1.0, the more accurate MRI was in distinguishing between benign and malignant ovarian tumors.

Figure 9 Sensitivity of studies: forest plot of sensitivities of 7 studies. Statistical method: inverse variance of the random-effects model. CI, confidence interval.

Figure 10 Specificity of studies: forest plot of specificities of 7 studies. Statistical method: inverse variance of the random-effects model. CI, confidence interval.

Figure 11 SROC curve for individual studies on MRI differential diagnosis of benign and malignant ovarian tumors. SROC, summary receiving operation characteristic; AUC, area under the curve; SE, standard error.

The positive likelihood ratio is the multiple of the correctly diagnosed disease and the wrongly diagnosed disease in the diagnostic auxiliary examination. Thus, in this study, the greater the positive likelihood ratio, the higher the accuracy of MRI in distinguishing between benign and malignant ovarian tumors. The analysis results showed that the pooled positive likelihood ratio of MRI in the differential diagnosis of benign and malignant ovarian tumors was 8.69 (95% CI: 5.06–14.92), indicating that MRI has a higher value for distinguishing between benign and malignant ovarian tumors (Figure 12).

Figure 12 Forest plot of positive LR. Comparison of positive LR between the benign group and the malignant group. Statistical method: inverse variance of the random-effects model. LR, Likelihood ratio; CI, confidence interval.

The negative likelihood ratio is the multiple of the probability of incorrectly diagnosing a negative disease, as the multiple of the negative probability of correctly diagnosing the disease. Thus, in this study, the smaller the negative likelihood ratio, the higher the accuracy of MRI as an auxiliary examination in distinguishing and diagnosing benign and malignant ovarian tumors. The negative likelihood ratio in this study was 0.07 (95% CI: 0.04–0.13), indicating that MRI is more accurate than CT in distinguishing between benign and malignant ovarian tumors (Figure 13).

Figure 13 Forest plot of negative LR. Comparison of negative LR between the benign group and the malignant group. Statistical method: inverse variance of the random-effects model. LR, Likelihood ratio; CI, confidence interval.

The diagnostic odds ratio reflects the closeness of the relationship between the diagnostic test results and the corresponding disease. The greater the diagnosis rate, the greater the ability to represent whether the disease is diagnosed by the diagnostic test or auxiliary examination. The results of this study showed that the combined diagnostic odds ratio of MRI for distinguishing between benign and malignant ovarian tumors was 146.19 (95% CI: 68.88–310.24). Thus, MRI is more accurate in distinguishing between benign and malignant ovarian tumors (Figure 14).

Figure 14 Forest plot of diagnostic odds ratio. Comparison of diagnostic odds ratio between the benign group and the malignant group. Statistical method: inverse variance of the random-effects model. OR, odds ratio; CI, confidence interval.

Risk of bias

The quality assessment found that in relation to the patient selection bias, 8 articles had a low risk (12-14,16-20) and 2 articles had an unclear risk (11,15). In the research index test bias, 3 articles had a high risk (11,16,17), 6 articles had a low risk (12,14,15,18-20), and 1 article had an unclear risk (13). In relation to the standard bias, flow and time bias, all the articles were low risk. In relation to the risk of applicability bias, the patient selection bias, index test bias, and reference standard bias all of the articles were low risk (Figure 15).

Figure 15 The intensity and distribution of the quality risk of the articles included in the study.

Discussion

The appearance of ovarian tumors is closely related to body factors and genetic factors, and is divided into benign tumors and malignant tumors. The former has a good prognosis after treatment. Malignant ovarian tumors are the third most gynecologically malignant tumor after cervical cancer and endometrial cancer. In the early stage of onset, patients lack typical symptoms, growth is fast, and there is a high rate of misdiagnosis and missed diagnosis. By the time of the pathological diagnosis, the tumor has often developed to the middle and late stages, which increases the complexity of the disease and the difficulty of treatment. Further, as the optimal treatment time has been missed, the life of the patient is even endangered. Thus, the diagnosis of the disease must be emphasized to improve the accuracy of the diagnosis results and provide a reference for the formulation and operation of disease treatment plans.

In recent years, due to rapid developments in imaging technology, the 64-slice spiral CT enhanced examination has come to be widely used in disease diagnosis. Due to its high efficiency, accuracy, and 3-dimensional advantages, it is used in the diagnosis of diseases in the circulatory system, respiratory system, digestive system, and various parts of the human body. The 64-slice spiral CT enables clinicians to obtain rich details, and also has a reduced radiation dose, and reduced layer thickness. It not only diagnoses disease morphology, but also analyzes benign and malignant tumors (21,22). MRI has many parameters and rich image information. It can scan different sections to display the anatomical structure and condition of the body under examination, which is conducive to the early detection of lesions. In addition, MRI has high soft tissue resolution and high qualitative diagnostic value (23,24). Thus, MRI plays an important role in the diagnosis and differential diagnosis of ovarian tumors.

According to the imaging characteristics, benign tumors usually have complete capsules and relatively regular shapes, while malignant tumors show aggressive growth, irregular shapes and incomplete capsules. Our analysis results show that both MRI and CT have good diagnostic value for ovarian tumors, but MRI has better sensitivity and specificity than CT, and a higher AUC value. Thus, MRI is better than CT in distinguishing between benign and malignant ovarian tumors. As CT scans only show the defect of the cross-sectional image, it is difficult to distinguish between endometriotic cysts, the uterine serosal layer or ovarian tumors (25,26). Conversely, MRI uses multi-directional and multi-level imaging, obtains a larger amount of information, a higher resolution of soft tissues, and clearly defines the range of edema, inflammation, tumors, etc., and thus provides certain biochemical and pathological information (21,24,26).

The current meta-analysis has some inherent limitations, including patient selection bias, research heterogeneity, and population differences. First, the number of articles included in the study was relatively small. Second, the articles included prospective and retrospective studies, which resulted in inconsistent data. Finally, only articles published in Chinese and English were included. Multi-center and large-sample studies need to be conducted to further clarify the sensitivity and predictive value of CT and MRI in the differential diagnosis of ovarian tumors.

Conclusions

For the differential diagnosis of benign and malignant ovarian tumors, MRI and CT have high diagnostic value. However, MRI had stronger differential diagnosis capabilities than CT and is a promising non-radioimaging technology, and thus may be a more advantageous choice for patients with ovarian tumors. In the future, large-sample, multi-center prospective studies are needed to weigh the value of MRI and CT in the diagnosis of ovarian tumors.

Acknowledgments

Funding: None.

Footnote

Reporting Checklist: The authors have completed the PRISMA reporting checklist. Available at https://gs.amegroups.com/article/view/10.21037/gs-21-889/rc

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://gs.amegroups.com/article/view/10.21037/gs-21-889/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Bray F, Ferlay J, Soerjomataram I, et al. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2018;68:394-424. [Crossref] [PubMed]
2. Kurman RJ, Shih IeM. The Dualistic Model of Ovarian Carcinogenesis: Revisited, Revised, and Expanded. Am J Pathol 2016;186:733-47. [Crossref] [PubMed]
3. Bowtell DD. The genesis and evolution of high-grade serous ovarian cancer. Nat Rev Cancer 2010;10:803-8. [Crossref] [PubMed]
4. Braem MG, Onland-Moret NC, van den Brandt PA, et al. Reproductive and hormonal factors in association with ovarian cancer in the Netherlands cohort study. Am J Epidemiol 2010;172:1181-9. [Crossref] [PubMed]
5. Gates MA, Rosner BA, Hecht JL, et al. Risk factors for epithelial ovarian cancer by histologic subtype. Am J Epidemiol 2010;171:45-53. [Crossref] [PubMed]
6. Moorman PG, Palmieri RT, Akushevich L, et al. Ovarian cancer risk factors in African-American and white women. Am J Epidemiol 2009;170:598-606. [Crossref] [PubMed]
7. Yang CY, Kuo HW, Chiu HF. Age at first birth, parity, and risk of death from ovarian cancer in Taiwan: a country of low incidence of ovarian cancer. Int J Gynecol Cancer 2007;17:32-6. [Crossref] [PubMed]
8. Ziogas A, Gildea M, Cohen P, et al. Cancer risk estimates for family members of a population-based family registry for breast and ovarian cancer. Cancer Epidemiol Biomarkers Prev 2000;9:103-11. [PubMed]
9. Mathieu KB, Bedi DG, Thrower SL, et al. Screening for ovarian cancer: imaging challenges and opportunities for improvement. Ultrasound Obstet Gynecol 2018;51:293-303. [Crossref] [PubMed]
10. Terry KL, Schock H, Fortner RT, et al. A Prospective Evaluation of Early Detection Biomarkers for Ovarian Cancer in the European EPIC Cohort. Clin Cancer Res 2016;22:4664-75. [Crossref] [PubMed]
11. Fan H, Wang TT, Ren G, et al. Characterization of tubo-ovarian abscess mimicking adnexal masses: Comparison between contrast-enhanced CT, 18F-FDG PET/CT and MRI. Taiwan J Obstet Gynecol 2018;57:40-6. [Crossref] [PubMed]
12. Guo L, Yang Z. Contrast-enhanced Ultrasound, CT and MRI in Ovarian Tumor Clinical Application. Chinese Journal of CT and MRI 2016;14:92-5.
13. Tsili AC, Tsampoulas C, Argyropoulou M, et al. Comparative evaluation of multidetector CT and MR imaging in the differentiation of adnexal masses. Eur Radiol 2008;18:1049-57. [Crossref] [PubMed]
14. Jiang M, Huang S, Sheng W, et al. Value of dual-energy CT in distinguishing between benign from malignant ovarian tumors. Hainan Med J 2020;31:3214-8.
15. Gity M, Parviz S, Saligheh Rad H, et al. Differentiation of Benign from Malignant Adnexal Masses by Dynamic Contrast-Enhanced MRI (DCE-MRI): Quantitative and Semi-quantitative analysis at 3-Tesla MRI Asian Pac J Cancer Prev 2019;20:1073-9. [Crossref] [PubMed]
16. Michielsen K, Dresen R, Vanslembrouck R, et al. Diagnostic value of whole body diffusion-weighted MRI compared to computed tomography for pre-operative assessment of patients suspected for ovarian cancer. Eur J Cancer 2017;83:88-98. [Crossref] [PubMed]
17. Pereira PN, Sarian LO, Yoshida A, et al. Accuracy of the ADNEX MR scoring system based on a simplified MRI protocol for the assessment of adnexal masses. Diagn Interv Radiol 2018;24:63-71. [Crossref] [PubMed]
18. Shimada K, Matsumoto K, Mimura T, et al. Ultrasound-based logistic regression model LR2 versus magnetic resonance imaging for discriminating between benign and malignant adnexal masses: a prospective study. Int J Clin Oncol 2018;23:514-21. [Crossref] [PubMed]
19. Thomassin-Naggara I, Poncelet E, Jalaguier-Coudray A, et al. Ovarian-Adnexal Reporting Data System Magnetic Resonance Imaging (O-RADS MRI) Score for Risk Stratification of Sonographically Indeterminate Adnexal Masses. JAMA Netw Open 2020;3:e1919896. [Crossref] [PubMed]
20. Zhang H, Mao Y, Chen X, et al. Magnetic resonance imaging radiomics in categorizing ovarian masses and predicting clinical outcome: a preliminary study. Eur Radiol 2019;29:3358-71. [Crossref] [PubMed]
21. Forstner R. Radiological staging of ovarian cancer: imaging findings and contribution of CT and MRI. Eur Radiol 2007;17:3223-35. [Crossref] [PubMed]
22. Gu P, Pan LL, Wu SQ, et al. CA 125, PET alone, PET-CT, CT and MRI in diagnosing recurrent ovarian carcinoma: a systematic review and meta-analysis. Eur J Radiol 2009;71:164-74. [Crossref] [PubMed]
23. Nam EJ, Yun MJ, Oh YT, et al. Diagnosis and staging of primary ovarian cancer: correlation between PET/CT, Doppler US, and CT or MRI. Gynecol Oncol 2010;116:389-94. [Crossref] [PubMed]
24. Rizzo S, Del Grande M, Manganaro L, et al. Imaging before cytoreductive surgery in advanced ovarian cancer patients. Int J Gynecol Cancer 2020;30:133-8. [Crossref] [PubMed]
25. Thomassin-Naggara I, Bazot M. MRI and CT-scan in presumed benign ovarian tumors. J Gynecol Obstet Biol Reprod (Paris) 2013;42:744-51. [Crossref] [PubMed]
26. Zhao S, Sun F, Bao L, et al. Pure dysgerminoma of the ovary: CT and MRI features with pathological correlation in 13 tumors. J Ovarian Res 2020;13:71. [Crossref] [PubMed]

(English Language Editor: L. Huleatt)