Introduction
Adnexal masses are among the most frequent diagnoses in gynecological practice with almost 20% of women developing a pelvic mass in their lifetime.1 When approaching an adnexal mass, the most feared diagnosis is ovarian cancer as it is one of the most lethal gynecologic malignancies primarily due to its late diagnosis, which can be attributed to the absence of an effective screening strategy and the lack of a symptomatic early phase. According to GLOBOCAN, in the year 2020, approximately 314,000 women were diagnosed with ovarian cancer, resulting in 207,000 deaths attributed to this disease.2 Furthermore, ovarian cancer is the eighth most prevalent cancer in terms of both incidence and mortality among women globally.2
Nonetheless, the majority of adnexal masses are benign conditions and that is why proper differentiation between malignant and benign lesions is crucial for adequate treatment. The preoperative diagnosis of adnexal masses is a matter of great relevance as it determines the management of the patients according to the risk of malignancy in terms of selecting the optimal surgeon (gynecologist-oncologist or general gynecologist) or surgical route (minimally invasive surgery or open surgery).3
Various strategies have been designed to provide clinicians with an accurate tool to determine whether the tumor is benign or ovarian cancer. It has been widely demonstrated that subjective assessment by an expert ultrasound examiner is considered the gold standard approach.4 However, when less experienced sonographers evaluate ovarian tumors, the use of IOTA simple rules or the ADNEX model yields comparable diagnostic performance.5
Despite significant advancements made in the field of ultrasound diagnosis for the majority of adnexal masses, this diagnostic tool still exhibits a considerable rate of false positive results, potentially resulting in a significant number of unnecessary procedures and heightened patient anxiety.6
Among these advancements, three-dimensional power-Doppler ultrasound was introduced into clinical practice in the late 90’s.7 Consequently, several research groups have assessed the potential role of tridimensional power Doppler ultrasound evaluation (3DPD) alongside the standard gray-scale morphologic ultrasonographic assessment of adnexal masses.
According to the literature, when using 3DPD, two different primary approaches for assessing an adnexal tumor have been proposed. One approach is based on the morphological characteristics of the tumor vascular tree and the second one evaluates the so-called three-dimensional vascular indexes, namely the vascularization index, flow index, and vascularization-flow index within the tumor.8
To the best of our knowledge, there is no meta-analysis analyzing the role of 3DPD in the differential diagnosis of adnexal masses. Such a meta-analysis would be valuable with potential scientific impact since it would analyze the current evidence about the role of this technique in assessing adnexal masses. In this study, we aimed to perform a systematic review and meta-analysis of 3DPD in the differential diagnosis of adnexal masses.
Methods
Search strategy
This systematic review was performed according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) recommendations (http://www.prisma-statement.org/ ). We did not register the protocol. Given the nature and design of this study, ethics committee approval was not required, and the study had no funding.
Three authors (AV, AC, ES) used three electronic databases (Web of Science, SCOPUS, and MEDLINE) [PubMed]) to identify potentially eligible articles that were published between January 1990 and May 2023. The search terms included the following keywords: “Ultrasound”, “Ovarian”, “Tumor” and “Three-dimensional”, and the search was limited to English language papers. One author (AV) combined the searches from the above-mentioned databases.
Duplicated articles and non-English articles were excluded. Subsequently, citations were screened first by the titles, then by the abstracts for identifying irrelevant articles to exclude (studies not related to the topic or not primary studies). Full-text articles of the remaining citations were read for the identification of potentially eligible papers. In studies from the same research group, we assessed the dates for recruitment, and in the case of overlap, we only considered the meta-analysis as the most recent study, unless they used different 3DPD approaches in different papers.
Two reviewers (JLA and AV) used the following criteria for selecting the articles: Prospective and retrospective cohort primary studies that include a set of patients who underwent 2D ultrasound evaluation and 3DPD in order to assess adnexal masses for discriminating between benign and malignant lesions and surgical evaluation of ovarian tumor for histopathological diagnosis as the reference standard. We excluded those articles that were not specifically related to the issue under review and studies that did not report data about morphological criteria or vascular indexes used for the adnexal mass evaluation. Any other studies not containing the necessary data to build a contingency table were also excluded. Three of the authors (AV, AC, and ES) gathered and were blinded from each other regarding data concerning the true positives, true negatives, false positives, and false negatives of each study. Any disagreement during this process was resolved by reaching a consensus among the three authors (AC, JLA, and ES).
Qualitative synthesis
The Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) was used to evaluate the risk of bias as well as concerns about the applicability of all studies included in this meta-analysis.9 This tool comprises four areas, namely “patient selection,” “index test,” “reference standard,” and “flow and timing.” Risk of bias and concerns about applicability were analyzed and rated as low, high, or unclear for each domain except that of flow and timing. The results of the quality assessment had a descriptive purpose in order to assess the global quality of the articles analyzed and to identify any potential factors of heterogeneity. The methodological quality was assessed independently by three authors (JLA, AC, and ES) using a standard form with quality assessment criteria and a flow chart. Disagreements were resolved by reaching a consensus among all three reviewers (JLA, AC, and ES).
The evaluation of the study’s quality was based on information such as study design, description of exclusion and inclusion criteria, and description of the index test and the reference standard test. For the index test, data regarding 3DPD evaluation methods was retrieved. Information on the diagnostic performance (true positives, true negatives, false positives, and false negatives) of 3DPD was also retrieved. This information was extracted separately for studies using 3DPD morphologic assessment of the tumor vascular tree and for studies using 3D vascular indexes.
Histopathological diagnosis was defined as the correct reference standard. To assess the flow-and-timing domain, we evaluated the description of the time elapsed between ultrasound examination and surgery (low risk of bias was considered when the reference standard was obtained less than 90 days after ultrasound evaluation).
Quantitative synthesis
We attempted to perform a quantitative synthesis including studies considered to be of moderate or high quality and that used comparable criteria in defining an adnexal mass as benign or malignant.
For that purpose, pooled specificity, sensitivity, and positive and negative likelihood ratio (LR) were determined using a random-effects model. As the estimation of 3D vascular indexes allows one to estimate the sensitivity and specificity of the method using three different indexes (namely vascularization index, flow index, and vascularization-flow index), we decided to use the vascularization index with the cut-off reported in each study.
Forest plots of sensitivity and specificity of all studies were calculated while heterogeneity for sensitivity and specificity was assessed using Cochran’s Q statistic and the I2 index.10 Summary receiver-operating characteristics (SROC) curves were plotted to illustrate the relationship between sensitivity and specificity. Publication bias was assessed according to Deek’s method.11 All analyses were performed using the MIDAS command in STATA (Stata Corporation, College Station, TX, United States) version 12.0 for Windows. Statistical significance was defined as a P-value of <0.05.
Results
Search results
The electronic search provided 404 citations, but after the exclusion of one hundred and forty-five duplicate records, 259 citations remained. Of these, one hundred and ninety-nine were excluded because it was clear from the title and/or abstract that they were not relevant to the review (papers not assessing diagnostic performance of 3DPD or not related to the topic).
Subsequently, the full texts of the 60 remaining articles were read. Finally, forty-three studies were excluded because they either did not assess diagnostic performance, were not related to the topic, were studied from the same group with overlapping recruiting dates, or a 2×2 table was not possible to obtain. The remaining 17 studies were ultimately included in the qualitative synthesis.12–28
A flowchart summarizing the literature search is shown in Figure 1.
Characteristics of the included studies
Seventeen studies reported from 2001 to 2021 were ultimately included, comprising 2,925 women with adnexal masses.12–28 Overall, 1,086 (37.0%) women had a malignant lesion. Table 1 provides a summary of the characteristics of the studies included in the present meta-analysis.12–28
Table 1Main characteristics of the studies included in the present meta-analysis
| Author | Year | Population | N patients | Patients’ age (years) | Postmenopausal patients | N malignant masses ( N BOT) | Study’s design | Consecutive series | Number of examiners | 3DPD approach | 3DPD criteria for suspicion | Reference standard |
|---|
| Kurjak12 | 2001 | Any mass | 292 | 54 (37–71)* | 34.9% | 30 (0) | N.A. | N.A. | One | Vascular tree | Vessel architecture disorganized and complex branching pattern | Histology |
| Cohen13 | 2001 | Complex masses | 71 | 22–80** | 43.7% | 14 (N.A.) | Prospective | N.A. | Three | Vascular tree | Presence of vessels in solid areas and/or septations | Histology |
| Alcazar14 | 2005 | Complex masses | 60 | 48 (17–82)† | 46.7% | 45 (4) | Retrospective | N.A. | One | Vascular tree | Presence of vessels in solid areas and/or septations | Histology |
| Geomini15 | 2006 | Any mass | 181 | 15–89** | 42.5% | 26 (11) | Prospective | Yes | Three | Vascular tree | Presence of vessels in solid areas and/or septations | Histology |
| | | | | | | | | | 3D VI whole tumor | N.A. | |
| Sladkevicius16 | 2007 | Any mass | 104 | N.A. | 32.1% | 21 (6) | N.A | Yes | One | Vascular tree | Vessels with abnormal branching, caliber changes, splashes and bridges. | Histology |
| Jokubkiene17 | 2007 | Any mass | 106 | N.A. | 41.5% | 21 (6) | Prospective | Yes | Multiple | 3D VI Sphere 5 cc most vascularized area | VI ≥ 10.6% | Histology |
| | | | | | | | | | 3D VI whole tumor | VI ≥ 2.26% | |
| Alcazar18 | 2008 | Complex masses | 39 | 48 (22–75)† | 43.6% | 20 (0) | Retrospective | Yes | One | Vascular tree | Vessels with irregular branching (>3 branches and close to 90° angulation branching), vessel caliber narrowing, microaneurysms, and vascular lakes | Histology |
| Dai19 | 2008 | Complex masses | 36 | 53 (19–91)† | 66.7% | 30 (5) | N.A. | N.A. | One | Vascular tree | Penetrating randomly dispersed vessels with ‘basket-like’ irregular branching | Histology |
| Chase20 | 2009 | Complex masses | 66 | 47 (18–77)* | N.A. | 10 (2) | N.A. | N.A. | One | Vascular tree | Chaotic flow pattern. Vessel sacculation. | Histology |
| Mansour21 | 2009 | Any mass | 400 | 11–83** | N.A. | 248 (0) | N.A. | Yes | One | Vascular tree | Chaotic pattern with complex distribution and branching. | Histology |
| Alcazar22 | 2009 | Complex masses | 143 | 50 (17–82)† | 53.8% | 113 (9) | Prospective | Yes | Two | 3D VI most vascularized solid area | VI ≥ 1.556% | Histology |
| Kudla23 | 2010 | Complex masses | 138 | 51 (18–88)† | 54.3% | 117 (7) | Prospective | Yes | Two | 3D VI Sphere 1 cc most vascularized area | VI ≥ 24.015% | Histology |
| | | | | | | | | | 3D VI Sphere 5 cc most vascularized area | VI ≥ 10.490 | |
| Perez-Medina24 | 2013 | Complex masses | 72 | 53 (22–86)† | 59.7% | 33 (8) | Prospective | Yes | One | Vascular tree | Vessel architecture disorganized and complex branching pattern | Histology |
| | | | | | | | | | 3D VI Sphere 5 cc most vascularized area | N.A. | |
| Silvestre25 | 2015 | Any mass | 75 | 18–82** | N.A. | 32 (5) | Prospective | Yes | One | 3D VI Sphere 4 cc most vascularized area | VI ≥ 3.4% | Histology |
| Utrilla-Layna26 | 2015 | Complex masses | 367 | 46 (18–80)† | 35.4% | 86 (4) | Prospective | Yes | One | 3D VI Sphere 1 cc most vascularized area | VI ≥ 24.015% | Histology |
| Smolen27 | 2016 | Any mass | 637 | N.A. | N.A. | 202 (N.A.) | N.A. | N.A. | N.A. | 3D VI not otherwise specified | N.A. | Histology |
| Sladkevicius28 | 2021 | Complex masses | 138 | 54 | 52.9% | 38 815) | Prospective | Yes | Multiple | Vascular tree | Vessels with abnormal branching, caliber changes, splashes and bridges. | Histology |
Study design was prospective in nine studies and retrospective in two studies.13–15,17,19,22–26,28 In six studies, study design was not reported.12,16,20–22,28 The series was consecutive in most studies.15–18,21–26,28
Six studies included any type of adnexal mass and 11 studies used 3DPD only in “complex” or “suspicious” masses on 2D gray-scale ultrasound.12–17,19–21,23,25,27,29
In 10 studies all 3DPD examinations were performed by the same single examiner,12,14,16,18–21,24–26 whereas in 6 studies more than one examiner participated in the 3DPD evaluation.13,15,17,22,23,28 In one study the number of examiners participating in the study was not reported.27
Nine studies used the assessment of the vascular tree as a criterion for discriminating between benign and malignant adnexal masses,12–14,16,18–21,28 six studies used the estimation of the 3D vascular indexes,17,22,23,25–27 and two studies assessed both approaches.15,24
Of those studies that used the assessment of the vascular tree as a diagnostic criterion, all reported the criteria used for considering the mass as suspicious, but the criteria were not the same for all studies (Table 1).
In the studies using the calculation of 3D vascular indexes, the methodologies employed were quite variable with one study including the whole tumor15; one including the whole tumor and an automated 5cc sphere of the most vascularized area of the tumor17; one including a manual estimation of the most vascularized area of the tumor22; one including two automated spheres, 1cc and 5cc, of the most vascularized area of the tumor23; one including an automated 5cc sphere of the most vascularized area of the tumor24; one including an automated 1cc sphere of the most vascularized area of the tumor26; another including an automated 4cc sphere of the most vascularized area of the tumor25; and, finally, one study did not report how estimation was performed.27
All studies used the histological diagnosis after surgical tumor removal as the reference standard.12–28 Nine studies reported on the time elapsed from ultrasound evaluation to surgery.12,15–17,19,24–26,28
Methodological quality of included studies
QUADAS-2 assessment of the risk of bias and concerns regarding the applicability of the selected studies is shown graphically in Figure 2.
Regarding the risk of bias in the domain “patient selection”, one study was not clear regarding patient inclusion criteria and eleven studies were considered as high risk because they included only selected masses for 3DPD assessment.12–14,18–20,22–24,26–28
Concerning the domain “index test”, sixteen studies adequately described the method as well as how it was performed and interpreted.12–26,28 One study did not describe how the 3DPD assessment was performed and was thus rated as unclear.27
For the domain “reference standard”, all studies were considered as low risk, since it was considered they correctly identified the target condition by the reference standard.12–28
Regarding the domain “flow and timing”, we considered nine studies low risk and the other eight unclear.12–28 Overall, the quality of the studies was considered good.
Concerning applicability, all studies were considered low risk for all three domains “patient selection”, “index test”, and “reference test”.
Diagnostic performance of 3DPD for discriminating between benign and malignant adnexal masses
Observing great heterogeneity in the methodologies used in the studies assessing tumor vascularization through the estimation of 3D vascular indexes, we decided not to perform a quantitative synthesis for this approach, since the results could not be compared.
Regarding the studies using the assessment of the features of the tumoral vascular tree, we observed that some studies included any type of mass while others included only “complex” or “suspicious” masses according to 2D gray-scale features. Therefore, we decided to analyze these studies separately.
Pooled sensitivity, specificity, positive likelihood ratio, and negative likelihood ratio of 3DPD vascular tree assessment for the four studies including any type of mass, comprising 936 women, were 77% (95% confidence interval [CI] = 52%–91%), 80% (95% CI = 37%–97%), 3.9 (95% CI = 0.7–20.9), and 0.29 (95% CI = 0.10–0.81), respectively. The diagnostic odds ratio was 14.0 (95% CI = 1.0–168.0). Significant heterogeneity for sensitivity (I2 = 95.7%, P < 0.001) and for specificity (I2 = 99.1%, P < 0.001) was found. Forest plots for sensitivity and specificity are shown in Figure 3. Meta-regression showed that none of the co-variables assessed as year of publication, sample size, and malignancy prevalence explained the heterogeneity observed. The area under the SROC curve for diagnostic performance of 3DPD in this group of lesions was 0.84 (95% CI = 0.81–0.87) (Fig. 4). Fagan’s nomogram shows that a 3DPD suspicious for malignancy in this group of lesions increases the pre-test probability from 30% to 63%; while a non-suspicious 3DPD decreases the pre-test probability from 30% to 11% (Fig. 5). We did not observe publication bias (P = 0.63).
Pooled sensitivity, specificity, positive likelihood ratio, and negative likelihood ratio of 3DPD vascular tree assessment for the seven studies including only “complex” or “suspicious” adnexal masses and comprising 493 women, were 90% (95% CI = 82%–94%), 88% (95% CI = 74%–95%), 7.3 (95% CI = 3.2–16.4), and 0.12 (95% CI = 0.06–0.22), respectively. The diagnostic odds ratio was 62.0 (95% CI = 17.0–217.0). Moderate heterogeneity for sensitivity (I2 = 54.9%, P = 0.04) and for specificity (I2 = 75.3%, P < 0.01) was found. Forest plots for sensitivity and specificity are shown in Figure 6. The area under the SROC curve for diagnostic performance of 3DPD in this group of lesions was 0.94 (95% CI = 0.92–0.96) (Fig. 7). Fagan’s nomogram shows that a 3DPD suspicious for malignancy in this group of lesions increases the pre-test probability from 45% (mean prevalence of malignancy in these studies was 45%) to 86% while a non-suspicious 3DPD decreases the pre-test probability from 45% to 9% (Fig. 8). We did not observe publication bias (P = 0.57).
Discussion
Summary of evidence
In this meta-analysis, we observed that 3DPD using the assessment of the tumor vascular tree had a good diagnostic performance for discriminating benign and malignant adnexal masses. The diagnostic performance was better when this technique was used in “complex” or “suspicious” adnexal masses. We also observed great heterogeneity in the methodological approaches of studies using the estimation of 3D vascular indexes as diagnostic criteria. In addition, we observed that the quality of the studies was moderate and there was room for improvement in study design and reporting.
Strengths and limitations
Strengths: This study is the first meta-analysis to address the issue, which is a significant strength.
Limitations: The study could not perform a quantitative synthesis for studies using the estimation of the 3D vascular indexes due to methodological differences among the studies. The quantitative synthesis for studies using the assessment of the tumor vascular tree was based on a limited number of studies and a small sample size, requiring caution in interpreting the results. The study did not compare 3DPD with 2D Color Doppler, which might affect the generalizability of the results. As high heterogeneity was observed among the studies, the results should be considered with caution.
Interpretation of the results in the clinical context
Adnexal masses are a common clinical problem in gynecological practice, and correct differential diagnosis is essential for adequate management. Currently, there is evidence that 2D gray-scale and color Doppler assessment of the adnexal masses, either by subjective examiner impression or using different classification systems, such as the IOTA Simple Rules, or predictive logistic models, such as the IOTA ADNEX model, are the best approach for discriminating between benign and malignant lesions.3,29
A meta-analysis showed that pooled sensitivity and specificity for the examiner’s subjective assessment was 90–94% and 85–94%, respectively.4 At least, three meta-analyses showed that pooled sensitivity and specificity for IOTA Simple Rules was 93–95% and 77–82%, respectively.4,30,31 Moreover, a recent meta-analysis observed that pooled sensitivity and specificity for the IOTA ADNEX model were 94% and 78%, respectively.32
In this context, the question is whether 3DPD adds diagnostic capacity to the ultrasound assessment of adnexal masses. According to our results, it seems that 3DPD does not add diagnostic information to current 2D ultrasound-based approaches for the differential diagnosis of adnexal masses, even in the selected populations. In addition, 3D ultrasound is not as widely available as 2D ultrasound. Furthermore, specific software is needed for assessing 3DPD findings in adnexal masses.