Optimizing Venographic Referrals in Pelvic Congestion Syndrome

Abstract

Objectives

Pelvic congestion syndrome (PCS) is a prevalent but underdiagnosed cause of chronic pelvic pain. Venography remains the diagnostic reference, yet its invasiveness limits routine use. This study evaluated standardized duplex ultrasound and developed a scoring system to guide selective, evidence‐based venographic referral.

Methods

In this prospective diagnostic‐accuracy study, 60 women with chronic pelvic pain underwent standardized transabdominal and transvaginal duplex ultrasound, and 40 subsequently underwent selective venography as the reference standard. Ultrasound parameters included venous diameter, evoked caudal flow duration, flow direction, velocity, bilaterality, and myometrial plexus enlargement. Penalized logistic regression with bootstrap‐nested cross‐validation identified independent predictors of venography‐confirmed PCS. Coefficients were converted into a point‐score to classify patients into rule‐out, indeterminate, and rule‐in bands. Diagnostic performance, calibration, and decision‐curve analyses were used to evaluate discrimination and clinical utility across probability thresholds.

Results

Venography confirmed PCS in 26 of 40 participants (65%). A 6‐mm diameter threshold provided optimal diagnostic balance (sensitivity 92.3%, specificity 64.3%). The penalized model retained diameter, reflux ≥1.0 second, low velocity ≤3 cm/second, Valsalva‐evoked caudal flow, myometrial plexus enlargement, and bilaterality, achieving an optimism‐corrected AUC of 0.861. The derived score stratified patients into rule‐out, indeterminate, and rule‐in groups with high predictive accuracy (positive predictive value 93.3%, negative predictive value 91.7%). Decision‐curve analysis showed higher net benefit than treat‐all or treat‐none strategies.

Conclusion

Standardized duplex ultrasound integrating morphologic and hemodynamic parameters accurately identified venographic PCS and enabled selective triage. The validated score supports ultrasound as an effective, noninvasive first‐line tool to optimize venography and guide patient‐centered management.

Abbreviations

AUC

area under the curve

BMI

body mass index

CI

confidence interval

IQR

interquartile range

LR

likelihood ratio

NPV

negative predictive value

NRS

numeric rating scale

PCS

pelvic congestion syndrome

PeVD

pelvic venous disorders

PPV

positive predictive value

SD

standard deviation

SMD

standardized mean difference

STARD

standards for reporting of diagnostic accuracy studies

TAS

transabdominal sonography

TR‐MRA

time‐resolved magnetic resonance angiography

TVS

transvaginal ultrasound

VO‐CPP

venous‐origin chronic pelvic pain

Chronic pelvic pain is a common and often disabling condition among reproductive‐age women, with vascular causes frequently overlooked. ¹ Pelvic congestion syndrome (PCS)—a term that largely overlaps with the more recently introduced concept of venous‐origin chronic pelvic pain (VO‐CPP) within the broader spectrum of pelvic venous disorders (PeVDs)—is characterized by venous valvular incompetence, venous dilatation, and reflux, and represents a treatable yet underdiagnosed contributor to chronic pelvic pain. ¹ , ² Venous pooling within ovarian and pelvic plexuses produces characteristic dull, positional pain that worsens with standing, exertion, or Valsalva. ¹ , ² Because symptoms overlap with various gynecologic and gastrointestinal conditions, PCS is often misclassified, delaying appropriate management. ³

Selective venography is the historical reference standard, offering direct assessment of reflux, vein diameter, and drainage patterns. However, its invasiveness, resource demands, and limited ethical indications make it impractical as a universal diagnostic test. ² In contrast, transvaginal and transabdominal ultrasound (TAS) provide a noninvasive, widely available modality capable of measuring venous caliber, flow velocity, and reflux under standardized physiologic maneuvers. ⁴ , ⁵ , ⁶ Although multiple studies demonstrate good correlation between ultrasound and venography, diagnostic thresholds remain inconsistent, with variable criteria for dilated veins (6–8 mm), sustained reflux (≥1 second), and low forward velocity (≤3 cm/second). ⁴ , ⁵ , ⁶ This lack of harmonized definitions has restricted reproducibility and slowed guideline adoption.

These discrepancies reflect broader methodological challenges: heterogeneous measurement techniques, differing visualization denominators, and inconsistent reference verification contribute to variable accuracy estimates. ⁷ , ⁸ Moreover, the clinical utility of ultrasound—particularly which features or thresholds best identify candidates for therapeutic venography—remains incompletely defined. As a result, decision‐making is fragmented, and unnecessary or delayed interventions persist. ⁷ , ⁸ , ⁹ A standardized, quantitative, and validated ultrasound‐based framework is urgently needed to optimize diagnostic efficiency and guide resource use.

This study aims to address these gaps by prospectively evaluating the diagnostic performance of combined transabdominal and transvaginal ultrasound across pelvic vein diameters above 4 mm and by developing and internally validating a multivariable predictive score for selective venography. By integrating anatomical, hemodynamic, and morphological ultrasound parameters into a penalized logistic model and translating coefficients into a practical bedside point score, this work seeks to establish ultrasound as an evidence‐based first‐line decision tool for PCS assessment.

Methods

Study Design and Setting

This prospective diagnostic‐accuracy study was conducted from December 2021 to April 2023 at 2 tertiary‐care hospitals—Firoozabadi and Firuzgar—affiliated with Iran University of Medical Sciences, Tehran, Iran. Consecutive women with chronic pelvic pain underwent standardized TAS and transvaginal (TVS) duplex ultrasound as the index test, and a clinically indicated subset proceeded to selective venography as the reference standard. Conduct and reporting followed Standards for Reporting of Diagnostic Accuracy Studies (STARD) 2015 guidelines. ¹⁰ The protocol was approved by the institutional review boards (IR.IUMS.FMD.REC.1402.016) and conducted in accordance with the Declaration of Helsinki; all participants provided written informed consent.

Participants

Consecutive women aged 18 years or older who presented with chronic pelvic pain lasting at least 6 months and were referred for pelvic ultrasound were screened for eligibility. Participants were included if initial sonographic assessment did not reveal an alternative cause for pain and if they could undergo both TAS and TVS scanning according to protocol. Exclusion criteria were sonographic or clinical evidence of conditions that could fully explain symptoms—such as large uterine fibroids, adenomyosis, endometriosis, pelvic inflammatory disease, pelvic tumors, portal hypertension, or major venous compression syndromes—as well as pregnancy, inability to provide informed consent, or contraindication to venography when clinically indicated. Screening, eligibility, and enrollment were recorded prospectively. Written informed consent was obtained for participation and for TVS when not part of the initial clinical order.

Ultrasound Protocol (Index Test)

All participants underwent standardized TAS and TVS duplex ultrasound performed by board‐certified attending radiologists using convex probes with frequencies optimized for each approach. Examinations were conducted in supine and semi‐erect positions, with spectral Doppler angle correction and physiologic maneuvers including coached Valsalva to elicit caudal flow. A predefined data dictionary guided acquisition of: (1) venous diameters of ovarian and parauterine veins as the mean of 3 repeated measurements of the maximal diameter; (2) reflux presence (duration ≥1.0 second threshold); (3) evoked caudal flow duration or least flow velocity (cm/second) at rest and during Valsalva; (4) arcuate or myometrial plexus enlargement; and (5) bilaterality of pelvic varices. Visualization adequacy and reasons for non‐visualization (eg, bowel gas, pain, or body habitus) were documented for feasibility analysis. Two experienced radiologists jointly calibrated technique on 10 pilot cases and were blinded to venography results. Quality control included random secondary review of cineloops and Doppler spectra for compliance with acquisition standards. Candidacy for venography was triaged by the ultrasound‐derived findings and patients were considered for venography only if they had pelvic or ovarian vein diameters ≥4 mm, and either (1) at least 1 additional abnormal ultrasound finding (eg, sustained reflux ≥1 second, forward velocity ≤3 cm/second, evoked caudal flow, bilaterality, or arcuate/myometrial plexus enlargement), or (2) typical clinical symptoms of PeVD that could not be explained by any other condition. ⁴ , ¹¹ , ¹² , ¹³ , ¹⁴

Reference Standard (Venography)

Selective catheter venography served as the reference standard for diagnosing PCS. Procedures were performed in an angiography suite under local anesthesia, using right common femoral or internal jugular venous access with 4–5 F catheters and nonionic contrast, injected manually or by power injector (10–20 mL per run at 3–6 mL/second). The left renal vein was catheterized, followed by selective catheterization of the ovarian vein using an angled catheter and hydrophilic guidewire. Venographic images were obtained at the vein origin during a Valsalva maneuver and/or with the table in a semierect position to best demonstrate venous distension and retrograde flow. The catheter was then advanced distally along the ovarian vein to delineate the distal gonadal vein anatomy, confirm retrograde flow, assess contrast wash‐out from distal veins and identify any connections supplying the pelvic venous plexus. The right ovarian vein was catheterized directly via the inferior vena cava and similarly assessed. Both internal iliac veins were also evaluated from their origins. Attention was paid to opacification of the reno‐caval and ilio‐caval segments to assess concomitant obstructive syndromes. A study was classified as positive when the following typical venographic findings were present: (1) evoked caudal (retrograde) flow of contrast in the ovarian or pelvic veins for ≥1 second during maneuvers, which was defined as venographic reflux, and/or (2) a combination of more than 1 of the congestion findings including an ovarian vein diameter more than 6 mm, opacification of ipsilateral or contralateral escape pathways (internal iliac tributaries, cross‐midline, vulvovaginal and thigh varicosities), and delayed contrast washout (≥20 second). ¹¹ , ¹² , ¹⁵

Vein calibers were measured at mid‐segments of mentioned veins on calibrated frames, and reflux column depth and drainage patterns were recorded. Venography readers were blinded to ultrasound results. When feasible, venography was performed with a priori therapeutic intent to allow same‐session embolization.

Ultrasound–Venography Concordance and Candidate Thresholds

Ultrasound and venography findings were compared in the paired subset of participants who underwent both tests. For each patient, the largest measured pelvic or ovarian vein diameter on ultrasound was matched to corresponding venographic measurements to evaluate concordance. Diagnostic performance was estimated for consecutive diameter thresholds (≥4, ≥5, ≥6, ≥7, and ≥8 mm), while reflux, flow velocity, and other categorical features were analyzed using predefined cutoffs (reflux ≥1.0 second; forward velocity ≤3 cm/second; presence of caudal flow, arcuate/myometrial plexus enlargement, or bilaterality). Agreement across modalities for continuous variables was examined using correlation and Bland–Altman methods. These paired data formed the analytic basis for evaluating single‐feature accuracy and for constructing the multivariable predictive model that combined anatomical, hemodynamic, and morphologic ultrasound features to predict venography‐confirmed PeVDs.

Treatment and Follow‐up Assessments

Patients with venography‐confirmed PCS were considered for transcatheter embolization during the same session, using detachable coils, vascular plugs, or liquid sclerosants according to vein size and anatomy. The goal was to occlude refluxing segments while preserving drainage through competent veins. Procedural success was defined by post‐embolization venography showing cessation of reflux, absence of collateral filling, and stasis within treated segments. Immediate and short‐term complications were prospectively recorded and graded by severity.

Follow‐up assessments were performed at approximately 3 months post‐procedure and included both clinical and sonographic evaluations. Pain intensity was measured on a numeric rating scale (NRS, 0–10), with treatment response defined as a reduction of ≥2 points. Follow‐up ultrasound measured largest pelvic vein diameter, peak venous velocity, evoked caudal flow duration, and arcuate/myometrial plexus enlargement. Changes from baseline were analyzed as paired mean differences with 95% confidence intervals (CI), and categorical features were evaluated using the McNemar test to determine concordant improvement in venous hemodynamics.

Blinding, Bias Mitigation, and Data Oversight

Ultrasound examinations were performed and recorded before knowledge of venography results to prevent interpretation bias. Venography readers were likewise blinded to ultrasound findings and to any intermediate modeling outputs. Both imaging teams used predefined case‐report forms and measurement dictionaries to ensure consistency across modalities. Spectrum and verification biases were mitigated by enrolling consecutive patients from routine clinical pathways and by transparently reporting visualization rates and denominators for all ultrasound features. Verification bias inherent to selective venography was addressed analytically through decision‐curve evaluation and stratified reporting of diagnostic performance within the paired subset. Internal validity was supported by model shrinkage, bootstrap‐based correction for optimism, and calibration diagnostics. Data were de‐identified, stored on secure institutional servers, and accessible only to authorized investigators. Study progress, protocol adherence, and safety outcomes were periodically reviewed by the research team to ensure integrity and compliance with ethical standards.

Statistical Analysis

Analyses were performed in R 4.3.2. Continuous variables were summarized as mean ± standard deviation (SD) or median (interquartile range [IQR]), and categorical variables as counts and percentages. Between‐group comparisons (PCS⁺ versus PCS⁻) used the t‐test or Wilcoxon rank‐sum test for continuous data and χ² or Fisher's exact test for categorical variables. For diagnostic evaluation, 2 × 2 contingency tables were constructed for each ultrasound feature and for consecutive vein‐diameter thresholds (≥4–≥8 mm). From these, sensitivity, specificity, positive and negative predictive values (PPV, NPV), likelihood ratios (LR⁺, LR⁻), and diagnostic odds ratios were estimated with 95% confidence intervals using nonparametric bootstrap resampling (B = 2000). Emphasis was placed on estimation and effect‐size interpretation rather than dichotomous hypothesis testing.

A penalized multivariable logistic regression model (elastic‐net) was developed to predict venography‐confirmed PCS using all ultrasound‐derived predictors—largest vein diameter, reflux, forward velocity, Valsalva‐evoked caudal flow, arcuate/myometrial plexus enlargement, bilaterality—and candidate covariates (age, body mass index [BMI], parity). Model tuning employed nested 10‐fold cross‐validation within 200 bootstrap resamples to estimate optimism‐corrected discrimination (area under the curve [AUC]), calibration slope/intercept, and Brier score. Shrinkage‐adjusted coefficients were linearly scaled into an integer point score, which was stratified into 3 diagnostic bands: rule‐out, indeterminate, and rule‐in, corresponding to high‐sensitivity/NPV and high‐specificity/PPV operating points. Decision‐curve analysis quantified the net benefit of score‐based triage compared with “treat‐all” and “treat‐none” strategies across plausible intervention thresholds. Missing data were imputed using multivariate chained equations (m = 20), and sensitivity analyses excluded suboptimal scans or long ultrasound‐to‐venography intervals to confirm robustness.

Results

Baseline Demographics and Clinical Characteristics

Sixty women were enrolled, including 40 who underwent venography (26 PCS⁺, 14 PCS⁻). The cohort had a mean age of 32.8 ± 6.1 years and mean BMI of 24.9 ± 3.2 kg/m², with no significant differences between diagnostic groups. Median parity was higher among PCS⁺ participants (2 [IQR: 2–3]) compared with PCS⁻ (1 [1, 2]; p = .04). Occupational standing exceeding 6 hours per day was more common in PCS⁺ cases (38.5% versus 21.4%; p = .03), while other lifestyle and gynecologic factors—including heavy lifting, hormonal therapy use, and prior pelvic surgery—did not differ meaningfully. Menopause and family history of PCS were infrequent. Baseline pain intensity was greater in PCS⁺ participants (7.1 ± 1.2 versus 6.1 ± 1.3; p = .01), representing a moderate‐to‐large effect size (standardized mean difference [SMD] = 0.81). Overall, groups were otherwise comparable in demographic and clinical characteristics (Table S1).

Ultrasound Feasibility and Vessel Visualization (TAS/TVS)

Ultrasound visualization was successfully achieved for most pelvic venous structures (Table S2). TVS consistently demonstrated higher visualization rates than TAS across all territories. The right and left ovarian veins were visualized in 86.7 and 90.0% of cases by TVS, compared with 73.3 and 70.0% by TAS, respectively. Internal iliac tributaries and uterine arcuate or myometrial plexus were also more frequently detected by TVS (80.0–91.7%) than TAS (61.7–63.3%). Visualization by both modalities was possible in approximately two‐thirds of participants, while complete non‐visualization occurred in 8–13% depending on structure. Among non‐visualized segments, the main limiting factors were body habitus, bowel gas, and pain intolerance during transvaginal examination. Overall, the combined TAS/TVS protocol provided high anatomic coverage suitable for quantitative analysis.

Reference Venography Findings (Gold Standard)

Among the 40 participants who underwent selective venography, PCS was confirmed in 26 (65.0%; 95% CI: 49.5–78.4) (Table S3). The largest pelvic vein diameter was greater in PCS⁺ compared with PCS⁻ participants (median 7.6 [IQR: 6.8–8.8] versus 5.1 [4.5–5.8] mm). PCS⁺ cases also exhibited a higher number of dilated veins (median 3 [2–4] versus 1 [0–1]) and a markedly higher prevalence of evoked caudal flow (92.3 versus 7.1%), with longer duration (1.4 [1.1–1.9] versus 0.4 [0.2–0.6] s); notably, no participant in the PCS‐negative group met the predefined criterion for reflux (caudal flow duration >1 second). In 2 PCS‐positive patients, no proximal reflux was observed on venography. However, in distal venography, prominent distal collateral veins and delayed contrast washout supported the diagnosis, indicating compensatory drainage and venous congestion despite the absence of visible reflux (due to inadequate patient cooperation to evoke reflux). Cross‐midline drainage was frequent among PCS⁺ patients (57.7%) but rare in PCS⁻ patients (7.1%), in whom it was observed solely as an isolated finding. Left‐sided or bilateral involvement predominated among PCS⁺ patients (15 left, 8 bilateral, 3 right). Collectively, these findings confirm the expected morphologic and hemodynamic distinctions between venography‐positive and negative participants.

Diameter Thresholds versus Venography (≥4 to ≥8 mm)

Diagnostic accuracy metrics for incremental ultrasound vein‐diameter thresholds are summarized in Table 1. Sensitivity was maximal at the lowest cutoff (≥4 mm: 100%) but accompanied by low specificity (28.6%). Raising the threshold to ≥5 mm improved specificity to 50.0% while maintaining high sensitivity (96.2%). Optimal balance was achieved at ≥6 mm, yielding sensitivity 92.3% and specificity 64.3%, with a diagnostic odds ratio of 21.6 (Figure 1). At higher cutoffs, specificity continued to rise—to 78.6% at ≥7 mm and 85.7% at ≥8 mm—while sensitivity gradually declined (84.6 and 76.9%, respectively) (Figure S1). PPV exceeded 80% for thresholds ≥6 mm, and the likelihood ratio for ≥8 mm (LR⁺ = 5.38) indicated moderate rule‐in utility. These results identify 6 mm as the most discriminative diameter threshold for venography‐confirmed PeVDs in this cohort.

Table 1.

Diameter Thresholds versus Venography (≥4–≥8 mm)

Diameter Cutoff (mm)	Sensitivity % (95% CI)	Specificity % (95% CI)	PPV % (95% CI)	NPV % (95% CI)	LR+	LR−	DOR
≥4	100.0 (87.1–100.0)	28.6 (11.7–54.6)	72.2 (56.0–84.2)	100.0 (51.0–100.0)	1.40	0.00	—
≥5	96.2 (81.1–99.3)	50.0 (26.8–73.2)	78.1 (61.2–89.0)	87.5 (52.9–97.8)	1.92	0.08	25.0
≥6	92.3 (75.9–97.9)	64.3 (38.8–83.7)	82.8 (65.5–92.4)	81.8 (52.3–94.9)	2.58	0.12	21.6
≥7	84.6 (66.5–93.9)	78.6 (52.4–92.4)	88.0 (70.0–95.8)	73.3 (48.0–89.1)	3.95	0.20	20.2
≥8	76.9 (57.9–89.0)	85.7 (60.1–96.0)	90.9 (72.2–97.5)	66.7 (43.7–83.7)	5.38	0.27	20.0

LR, likelihood ratio; DOR, diagnostic odds ratio; PPV, positive predictive value; NPV, negative predictive value.

Figure 1.

Diagnostic operating points across diameter thresholds (≥4–≥8 mm): sensitivity and specificity. Lines display sensitivity and specificity (%) for each pre‐specified caliber threshold using the venography‐paired subset (N = 40). Points are annotated at each threshold.

Penalized logistic regression for venography‐confirmed PCS multivariable modeling with elastic‐net penalized logistic regression identified several ultrasound‐derived predictors independently associated with venography‐confirmed PCS (Table 2). The largest pelvic vein diameter (per +1 mm above 4 mm) showed the strongest linear effect (β = 0.34; OR: 1.41, 95% CI: 1.10–1.97), retained in 90.8% of bootstrap resamples. Additional ultrasound‐derived predictors included Valsalva‐evoked caudal flow (β = 1.17; OR: 3.22, 95% CI: 1.28–8.86) and reflux duration ≥1.0 second (β = 0.83; OR: 2.30, 95% CI: 1.01–5.61). Lower‐velocity flow (≤3 cm/second), arcuate/myometrial plexus enlargement, and bilateral dilation demonstrated moderate contributions (retention 71–76%). Age contributed weakly, while BMI and parity were near‐zero after shrinkage. Internal validation yielded an apparent AUC of 0.882 and optimism‐corrected AUC of 0.861, with good calibration (slope 0.93, intercept 0.04) and a Brier score of 0.158. These results indicate robust model discrimination and reliability for ultrasound‐based prediction of venographic confirmed PCS.

Table 2.

Multivariable Elastic‐Net Model for Ultrasound‐Based Prediction of Venography‐Confirmed PCS (Shrinkage‐Adjusted)

Predictor (Definition/Unit)	β (Shrinkage‐Adjusted)	OR (95% CI)	Retained (%)
Largest diameter (per +1 mm above 4)	0.34	1.41 (1.10–1.97)	90.8
Evoked caudal flow (present vs absent)	1.17	3.22 (1.28–8.86)	89.1
Reflux (duration ≥1.0 second) (yes vs no)	0.83	2.30 (1.01–5.61)	82.0
Low velocity ≤3 cm/second (yes vs no)	0.78	2.18 (0.95–4.96)	76.0
Arcuate/myometrial plexus enlargement (present vs absent)	0.62	1.86 (0.84–4.20)	72.9
Bilaterality of dilation (yes vs no)	0.53	1.70 (0.72–3.92)	71.3
Age (per −5 years)	0.21	1.23 (0.90–1.76)	59.6
BMI (per kg/m2)	0.02	1.02 (0.94–1.11)	28.4
Parity (per birth)	0.01	1.01 (0.89–1.16)	24.1

BMI, body mass index; OR, odds ratio; PCS, pelvic congestion syndrome.

Integer Point‐Score Derived from Penalized Coefficients

Shrinkage‐adjusted logistic coefficients were converted into an integer point‐score to facilitate bedside application (Table 3). Weighted contributions were assigned as follows: +2 per each +1 mm increase above 4 mm for vein diameter, +4 for presence of evoked caudal flow, +3 for reflux ≥1.0 second, and +2 each for low forward velocity (≤3 cm/second), arcuate/myometrial plexus enlargement, and bilaterality of dilation. Age < 35 years contributed +1 point. The resulting total score spanned 0–16 points. Calibration of score‐to‐probability mapping demonstrated a monotonic increase in venography‐confirmed PCS likelihood—from approximately 12% at 0–1 points to >95% at ≥15 points (Figure 2). Intermediate bands reflected graded post‐test probabilities of ~20, ~35, ~55, ~70, and ~85% for increasing score intervals, supporting its interpretability for clinical triage and decision‐making.

Table 3.

Integer Point‐Score Derived from Penalized Coefficients

Predictor (Coding)	Assigned Points
Largest diameter (per +1 mm above 4 mm)	+2
Evoked caudal flow present	+4
Reflux (duration ≥1.0 second)	+3
Low velocity ≤3 cm/second	+2
Arcuate/myometrial plexus enlargement present	+2
Bilaterality of dilation	+2
Age <35 years	+1

Figure 2.

Total score to estimated probability mapping (step/band curve). The x‐axis is the summed points; the y‐axis is the corresponding estimated probability of venography‐positive PCS. Discrete steps are shown for score ranges with a single mapped probability, and a shaded band is used where the model supports a probability interval.

Three‐Band Classification (Rule‐out/Indeterminate/Rule‐in)

Application of 3 diagnostic bands to the ultrasound‐derived point‐score provided a pragmatic framework for clinical triage (Table 4). The rule‐in band (≥10 points) identified 15 of 40 participants (37.5%), including 14 PCS⁺ and 1 PCS⁻, yielding a within‐band PPV of 93.3% (95% CI: 70.2–98.8). As a binary classifier, this cutoff provided specificity of 92.9% and LR⁺ of 7.54, indicating strong rule‐in capability. The indeterminate band (4–9 points) encompassed 13 participants (32.5%) with a PCS prevalence of 84.6% (57.8–95.7), representing a diagnostic gray zone warranting confirmatory testing. The rule‐out band (≤3 points) comprised 12 participants (30.0%) with only 1 PCS⁺ case, corresponding to NPV of 91.7% (64.6–98.5) and LR⁻ of 0.09 when interpreted inversely. Collectively, this 3‐band system balanced triage efficiency by distinguishing low‐ and high‐probability groups while minimizing false classifications in intermediate scores.

Table 4.

Three‐Band Classification Using the Point‐Score

Band (Score Range)	n (%)	PCS+, n (%)	PCS−, n (%)	PCS Prevalence % (95% CI)	As Binary Test for PCS		LR+	LR−
					Sensitivity %	Specificity %
Rule‐in (≥10)	15 (37.5)	14 (93.3)	1 (6.7)	93.3 (70.2–98.8)	53.8	92.9	7.54	0.50
Indeterminate (4–9)	13 (32.5)	11 (84.6)	2 (15.4)	84.6 (57.8–95.7)	42.3	85.7	2.96	0.67
Rule‐out (≤3)	12 (30.0)	1 (8.3)	11 (91.7)	8.3 (1.5–35.4)	3.8	21.4	0.05	4.49

PCS, pelvic congestion syndrome; LR, likelihood ratio.

Score‐Based Operating Characteristics at Common Clinical Thresholds

Score‐based classification demonstrated robust diagnostic performance at conventional clinical thresholds (Figure 3, Table S4). Using a rule‐in threshold (≥10 points), 14 of 15 positive cases were true positives, producing a PPV of 93.3% (95% CI: 70.2–98.8), specificity 92.9%, and LR⁺ 7.54, indicating strong discriminatory power for confirming PCS. Applying a rule‐out threshold (≤3 points) yielded sensitivity 96.2% and LR⁻ 0.18, with NPV 75.0% (42.8–92.3), effectively minimizing false negatives but at lower specificity (21.4%). When both cutoffs were applied for 2‐sided triage, 27 of 40 cases (67.5%) were classified: 14 true positives, 1 false positive, 1 false negative, and 11 true negatives, corresponding to PPV 93.3% and NPV 91.7% (Figure S2). The remaining indeterminate band (13/40; 32.5%) represented cases requiring shared clinical decision‐making or confirmatory imaging. This approach supports practical ultrasound‐based stratification for venography referral.

Figure 3.

Global performance comparison across strategies (radar plot). Normalized polygons compare sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), likelihood‐ratio (LR+), and (1/LR−) for rule‐in only, rule‐out only, and 2‐sided triage. Likelihood‐ratio axes are range‐normalized to allow commensurate radial scaling with probability metrics.

Internal Validation and Decision‐Curve Metrics

Model performance after internal validation demonstrated strong discrimination and calibration. Bootstrap‐nested 10‐fold cross‐validation yielded an optimism‐corrected AUC of 0.861, calibration slope 0.93, calibration intercept 0.04, and Brier score 0.158, confirming reliable predictive accuracy and minimal overfitting. Decision‐curve analysis further quantified the net clinical utility of the score‐based strategy compared with “treat‐all” and “treat‐none” approaches (Table S5). Across a range of threshold probabilities (pt = 0.20–0.60), the point‐score consistently offered the highest net benefit, ranging from 0.352 at pt = 0.20 to 0.102 at pt = 0.60, outperforming the treat‐all strategy by absolute margins of 0.05–0.07. Corresponding gains translated to an estimated 18–26 avoided venographies per 100 patients, without increasing false‐negative risk. These findings indicate that the ultrasound‐derived score maintains favorable clinical benefit over a wide range of decision thresholds, supporting its role as a selective, evidence‐based triage tool for venography referral.

Clinical and Sonographic Outcomes at Follow‐Up

Among treated participants with available follow‐up (n = 18/20, 90%) after a median of 90 days [IQR 75–120], both clinical and sonographic indices showed significant improvement (Table 5). Mean pain intensity (NRS, 0–10) declined from 7.1 ± 1.2 at baseline to 2.3 ± 1.5 at follow‐up (Δ = −4.8, 95% CI: −5.5 to −4.1; p < .001). The standardized response magnitude was large (SRM = 2.9), and 16 of 18 patients (88.9%) achieved a ≥ 2‐point reduction. Corresponding sonographic parameters demonstrated concordant hemodynamic recovery: largest vein diameter decreased from 7.8 ± 1.0 to 5.9 ± 1.1 mm (Δ = −1.9 mm; p < .001), peak venous velocity increased from 2.7 ± 0.6 to 3.5 ± 0.7 cm/second (Δ = +0.8 cm/second; p < .001), and evoked caudal flow duration shortened from 1.2 ± 0.4 to 0.6 ± 0.3 second (Δ = −0.6 second; p < .001). The prevalence of arcuate/myometrial plexus enlargement declined from 67% to 28% (p = .02), indicating both symptomatic and imaging‐based improvement following intervention.

Table 5.

Clinical and Sonographic Outcomes at Follow‐up (Within‐Patient Change)

Outcome	Baseline	Follow‐up	Δ (95% CI)	Paired p	SRM/Cohen's dz	Missing, n (%)
Pain intensity (NRS 0–10), mean ± SD	7.1 ± 1.2	2.3 ± 1.5	−4.8 (−5.5 to −4.1)	<.001	2.9	0 (0)
Largest vein diameter, mean ± SD, mm	7.8 ± 1.0	5.9 ± 1.1	−1.9 (−2.4 to −1.4)	<.001	1.7	1 (5.6)
Peak venous velocity, mean ± SD, cm/second	2.7 ± 0.6	3.5 ± 0.7	+0.8 (+0.5 to +1.1)	<.001	1.3	1 (5.6)
Evoked caudal flow duration, mean ± SD, seconds	1.2 ± 0.4	0.6 ± 0.3	−0.6 (−0.8 to −0.4)	<.001	1.5	2 (11.1)
Arcuate enlargement*, n (%)	12 (67%)	5 (28%)	—	.02	—	0 (0)

NRS, numeric rating scale; SD, standard deviation.

Discussion

The present prospective diagnostic‐accuracy study demonstrated that ultrasound, when performed with a standardized protocol and quantified across multiple anatomic and hemodynamic parameters, offers high discriminative ability for identifying venous reflux and dilation consistent with PCS. A 6 mm diameter threshold provided the most balanced sensitivity and specificity, and incorporation of additional features—such as evoked caudal flow, reflux duration ≥1.0 second, and myometrial plexus enlargement—further enhanced diagnostic performance. The resulting penalized multivariable logistic model achieved excellent internal validation (AUC = 0.861) and was translated into a practical bedside point‐score that stratified patients into rule‐out, indeterminate, and rule‐in categories with high predictive accuracy. Decision‐curve analysis confirmed meaningful clinical benefit by reducing unnecessary venography without increasing false negatives. Collectively, these findings support the use of duplex ultrasound as a reliable first‐line triage tool in women presenting with chronic pelvic pain suggestive of venous congestion.

These findings align with early sonographic studies of pelvic venous insufficiency, which highlighted the combined evaluation of venous caliber, reflux, and myometrial plexus involvement, demonstrating that ovarian and parauterine vein dilation with Valsalva‐induced caudal flow could be reliably identified in symptomatic women. ¹³ Standardized scanning maps for PCS later formalized these assessment targets and maneuvers, likely contributing to our high visualization rates and reproducible hemodynamic data, with our acquisition protocol closely reflecting those established frameworks. ¹⁶ The stepwise change in diagnostic accuracy across 4–8 mm thresholds supports prior consensus that pelvic veins under 4 mm are physiologic, while larger diameters increase suspicion but require contextual interpretation. ¹² Studies using TVS often identify ≥8 mm as a specific rule‐in marker, aligning with our observed high specificity and PPV at higher thresholds. ⁴ However, consistent with previous duplex and cross‐sectional imaging reports, our findings emphasize that diameter alone is insufficient for diagnosing reflux, as physiologic dilatation and false positives may occur without corroborating hemodynamic evidence. ¹⁷ , ¹⁸ Integration of hemodynamic parameters, particularly sustained reflux, low forward velocity, and Valsalva‐evoked caudal flow, improved discrimination in our cohort. These findings are consistent with recommendations to combine structural and functional evidence. ¹² , ¹³ Myometrial plexus enlargement and bilaterality also contributed to discrimination, in line with prior observational findings. ⁴

Multivariable modeling is increasingly recognized as essential for diagnostic decision‐making in PCS. A recent TVS‐based model showed that combining caliber, reflux, and plexus features enhances discrimination compared with single parameters. ¹⁹ Our penalized logistic framework extends this approach, incorporating regularization, bootstrap‐nested cross‐validation, and calibration diagnostics, as recommended for clinical prediction research. ²⁰ , ²¹ The resulting discrimination and calibration metrics parallel those reported in other well‐specified ultrasound prediction models for venography‐confirmed PCS. ¹⁹ Transforming regression coefficients into a bedside score with rule‐out, indeterminate, and rule‐in bands reinforces ultrasound's role as a selective venography gatekeeper. ¹¹ , ¹⁴ Concentrated post‐test probabilities and preserved NPV support minimizing unnecessary procedures, while decision‐curve analysis quantifies clinical utility—an aspect seldom addressed in PCS imaging studies. ²² Using venography as the diagnostic reference aligns with current consensus ¹² ; however, time‐resolved magnetic resonance angiography (TR‐MRA) has shown strong concordance for detecting and grading ovarian reflux, offering a useful noninvasive adjunct. ²³ Our ultrasound‐based probability stratification complements this framework by identifying indeterminate cases that may benefit most from targeted TR‐MRA evaluation. Clinical outcomes after endovascular therapy in our cohort, including high technical success, low complication rates, and marked early pain relief with parallel hemodynamic improvement, align with prior evidence showing durable benefit and safety after targeted embolization of refluxing pelvic veins. The concordance between imaging‐defined hemodynamics and clinical response reinforces ultrasound‐guided triage and underscores the central role of targeted embolization in achieving sustained therapeutic success. ²⁴ , ²⁵

The present findings clarify the physiologic and clinical meaning of duplex markers by linking reflux duration, flow direction, and venous caliber to the underlying failure of valvular competence. ¹² , ¹³ Combining these parameters with myometrial and bilateral venous involvement supports a mechanistic rather than purely morphometric view of pelvic congestion. ⁴ The derived score provides a structured triage pathway: low‐probability results justify observation, high‐probability results prioritize targeted venography with therapeutic intent, and indeterminate results prompt adjunct imaging such as TR‐MRA. ²³ This framework advances ultrasound from a descriptive modality to a decision‐support tool that can guide personalized investigation strategies, reduce unnecessary invasiveness, and align resource use with clinical need. ¹¹ , ¹⁴ Its transparent thresholds also facilitate shared decision‐making, enabling patient participation in balancing procedural risk against potential benefit while maintaining ethical and stewardship standards. ²⁶

Strengths of this study include its prospective design, standardized dual‐modality ultrasound protocol, blinded reference verification, and adherence to modern prediction‐modeling guidelines. ²¹ , ²⁶ Internal validation through bootstrap‐nested cross‐validation and calibration diagnostics enhances confidence in model robustness, while inclusion of decision‐curve analysis directly links statistical performance to clinical utility. ²² However, limitations must be acknowledged. The moderate sample size and referral‐based spectrum may overrepresent advanced disease, limiting external generalizability. Conditional verification could introduce bias, as venography was mainly performed in higher‐suspicion cases. External validation and recalibration in new cohorts are required to confirm reproducibility and adjust probability mapping to local prevalence. Follow‐up was limited to early post‐embolization outcomes, precluding evaluation of long‐term durability. Finally, reproducibility across operators and systems was not tested, emphasizing the need for structured training and performance monitoring. ¹⁶

Future work should prioritize multicenter external validation and recalibration of the score across diverse populations and practice settings. Prospective impact studies are needed to verify whether score‐based triage reduces unnecessary venography while maintaining or improving patient‐centered outcomes, such as pain relief, function, and quality‐adjusted life‐years. Broader adoption will benefit from harmonized definitions of outcomes and standardized follow‐up intervals, supporting development of a core outcome set for PeVDs. ²⁷ Advances in duplex imaging, integration with TR‐MRA, and application of AI‐driven flow analysis may further refine this ultrasound‐first diagnostic strategy for PCS.

This study demonstrates that standardized duplex ultrasound accurately identifies PCS and guides selective venography. A 6‐mm diameter threshold combined with hemodynamic features achieved strong diagnostic performance, and the derived point‐score enabled practical triage into rule‐out, indeterminate, and rule‐in categories. Concordant clinical and imaging improvements after targeted embolization support ultrasound as an effective first‐line diagnostic tool, though larger multicenter studies are needed to confirm generalizability and long‐term outcomes.

Supporting information

Figure S1. Confusion‐matrix tiles per diameter threshold (2 × 2 counts). For each threshold (≥4 to ≥8 mm), a mini heatmap displays True Positive/False Negative in the top row (actual disease positive) and False Positive/True Negative in the bottom row (actual disease negative). Counts are taken directly from Table 4 (paired subset N = 40).

Figure S2. Confusion‐matrix panels for score‐based strategies (three 2 × 2 heatmaps). Each panel depicts Actual+/Actual− by Predicted+/Predicted− for (left → right): rule‐in only, rule‐out only, and 2‐sided triage strategies. Cell labels show raw counts (true positive, false negative, false positive, true negative) for the venography subset, with text outlined for legibility.

Table S1. Baseline demographics and clinical characteristics.

Table S2. Ultrasound feasibility and vessel visualization (TAS/TVS).

Table S3. Reference venography findings (gold standard).

Table S4. Score‐based operating characteristics at common clinical thresholds.

Table S5. Decision‐curve metrics.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.