The relationship between pelvic vein incompetence and chronic pelvic pain in women: systematic reviews of diagnosis and treatment effectiveness

Clinical symptoms specific to pelvic congestion syndrome

As CPP can comprise many symptoms, there is considerable overlap between those whose pain can be attributed to organic pathologies such as endometriosis and those with pain associated with PVI. When women with PCS were compared with women with CPP due to other pathology, the former were more often multiparous, and would describe the pain as on one side of the abdomen, dull and achy with sharp exacerbations, made worse by long periods of standing and walking. Conversely, lying down relieved the symptoms. Abdominal palpation over the ovarian point and a history of post-coital ache are discriminatory factors for pelvic congestion rather than for any other causes of pelvic pain (94% sensitivity and 77% specificity). 6 Visible, disfiguring vulval varicosities are also often present, although these are reported to temporarily occur in 2–20% of pregnancies;11 there are no data on their prevalence in non-pregnant women.

Pelvic congestion syndrome is considered by many to be a condition of the reproductive years, and most women with a PCS diagnosis are pre menopausal, suggesting that female reproductive hormones may influence the condition. However, some case series of embolisation of pelvic veins report treatment on women up to 75 years of age. Many cases of pelvic varices are found incidentally during Doppler ultrasound screening of the lower limb in women with varicose veins, and vascular specialists believe that treatment is warrented to prevent a recurrence of varicose veins, irrespective of pain.

A higher level of anxiety and depression is noted in women with CPP. 5 Pain is a significant risk factor for psychological distress, and, for women without a explaination for the their pain, this will be exacerbated. A tentative physiological mechanism is that the stretching and shearing forces of refluxing blood flow on the endothelial and smooth muscle cells of the pelvic vein prompt the release of vasodilators. These include neuropeptide transmitters such as substance P and neurokinins A and B, which are integral to the neurological pathways involved in stress and the regulation of emotions.

Other distinguishing features of women with PCS, compared with controls, are larger uteruses and thicker endometria. Over half of women (56%) with PCS have been found, on ultrasound scans, to have cystic changes on their ovaries, ranging from classic polycystic patterns to clusters of 4–6 cysts,12 but not nessarily concurrent with uterine hypertrophy. These observations suggest that ovarian dysfunction plays a role in the pathogenesis of PCS, but is not the sole factor.

Risk factors

A relationship with pregnancy was noted early. The observation that the OV capacity of a pregnant women at term is 60 times that of a non-pregnant woman13 led to the presumption that this contributed to chronic distension post partum. Around one-third of women will develop venous insufficiency during their first pregnancy, with the prevalence of varicose veins rising with each successive pregnancy. 14 Several mechanisms are postulated, including the increased volume of blood volume, the mechanical compression by the gravid uterus on pelvic veins and inferior vena cava and increase in progesterone that cause a reduction in venous tone. Varicose veins of the leg tend to improve in the months after delivery, but little is known about the natural history of pelvic varices.

Contradictory data exist, however, with Kim et al. 15 reporting that 63% of participants with PVI were nulliparous, and no differences in symptomatic response between these and parous women were seen following embolisation. The landmark paper by Beard et al. 16 found no difference in a composite measure of PVI between parous and nulliparous women. PCS in nulliparous women was attributed to disturbances of the autonomic nervous system, owing to the observation that vaginal wall blood flow increased when a woman was anxious and decreased when she was relaxed. 17 This corroborated the prevailing opinion at that time that pelvic congestion symptoms were at least partially psychological. This neglected the possibility that enduring chronic pain might cause psychological symptoms such as depression and anxiety, rather than being a manifestation of them, as discussed above. Although there appears to be an excess of parous women in many cohorts studied, parity alone cannot be a determining risk factor.

Nutcracker syndrome

One particular anatomical variant is observed when the left renal vein is compressed between the superior mesenteric artery and the aorta. This creates hypertension in the left renal vein and may lead to the development of new collateral connecting veins around the renal pelvis, but also, if it persists, it can lead to renal vein incompetence and massive reflux into the left OV and the development of OV varices. 21 The symptoms are different from those of PCS in that pain is concentrated on the left flank, haematuria is present and images show not only pelvic and vulval varices but an elevated pressure gradient between the left renal vein and vena cava. The condition is improved by endovascular stenting.

Pathophysiology

The cause of the dilated veins is poorly understood but is thought to be a consequence of both mechanical stresses and ovarian dysfunction.

The pelvic veins are known to dilate during pregnancy, by an estimated 60 times their normal diameter, to accommodate the increase demand on the vascular system. These changes generally regress over around 6 months, but they may persist and then become compounded by subsequent pregnancies. In pregnancy, 20% of women will develop varicose veins in the lower limbs but only one-third of these women will have vulvar varices. 14 It is proposed that compression of pelvic veins by the gravid uterus against the vertebrae leads to PVI, but the complex web of veins in the pelvis allows redirection of the flow. The valves in superficial veins (perineal and labial) do not allow reflux from the pelvis, but again in pregnancy these valves too become incompetent and provoke the development of varices in the vulva, perineum and lower limbs.

Oestrogen is a potent pelvic vasodilator via nitric oxide release, smooth muscle relaxation and loss of vascular responsiveness. 22 Polycystic ovaries are frequently seen in excess in women with PCS. 23 It has been observed that counteracting this with intravenous synthetic ergotamine reduced pain and improved venous flow in the pelvic veins. 24 Temporary down-regulation of ovarian function by medroxyprogesterone acetate (MPA), progestin or gonadotropin-releasing hormone (GnRH) agonists,10,25,26 (see Treatment), improves pain, again implicating an endocrine role in PVI. The regression of the condition after menopause corroborates this theory.

Venography

Venography is the real-time imaging of veins by radiography following the injection of a contrast agent. In earlier studies, the contrast was injected 0.5–1 cm into the myometrium under fluoroscopic visualisation and the diffusion into the venous system observed by taking images at 20-second intervals thereafter. This has been universally superseded by transcatheter venography, whereby a catheter is introduced into the venous system, usually via the common femoral or jugular vein, and guided to the chosen vein. The contrast agent is injected directly and its diffusion observed in real time. Radiologists assess the venous diameter, retrograde (reversed) blood flow, the presence of redundant collateral veins between veins and the delayed or stagnant clearance of contrast at the end of injection. Venography is usually performed on the left and right OVs and left and right IIVs, but some radiologists also selectively catheterise the renal veins and the external iliac veins.

Although the venogram study does require radiation and the use of contrast, and is invasive, it has several advantages over other imaging and so is considered the definitive method of identifying PVI. The diagnostic venogram gives immediate dynamic flow information and measurements of pelvic veins with the option of tilting the patient position. Furthermore, the patient is conscious and so can be asked to perform the Valsalva manoeuvre to further delineate venous reflux. This involves asking the patient to exhale against a closed glottis, for example by pinching her nose closed. This increases the intrathoracic pressure and ultimately the venous return to the heart, increasing peripheral venous pressures. The Valsalva manoeuvre prompts a short and limited reflux in competent veins, while in incompetent veins it provokes a pronounced and long-lasting reflux. If PVI is identified, the treating interventional radiologist can immediately proceed to embolisation, if this is so desired.

Ultrasound

Ultrasound uses reflected sound waves of varying frequencies to generate images. Higher frequencies produce a higher-resolution image but do not penetrate as deeply into tissues and hence transabdominal ultrasound (TAUS) uses transducers of 3–5 MHz, while transvaginal ultrasound (TVUS), being closer to the organs, can use higher-frequency transducers and, therefore, produce a better image. Standard ultrasound can identify venous structures but, to assess blood flow, Doppler ultrasound is required. Ultrasound bounces off the blood cells in a vein, causing a change in pitch of the reflected sound waves (called the Doppler effect). If there is no blood flow, the pitch does not change. Information from the reflected sound waves is processed to provide graphs or superimposed zones on the anatomical image that represent the flow of blood by the amplitude or brightness of the image.

Conventional Doppler technique relays two aspects: the velocity of the blood flow as peaks on a timeline and power of the signal on a grey scale. Colour Doppler expresses velocity and direction together, with red peaks indicating a positive shift and blue peaks indicating a negative one, with a superimposed white image reflecting increasing speed. Power Doppler shows only the strength of the Doppler signal, with the intensity of the white image proportional to the power, and omits speed and direction from the output. This renders the position of the transducer less important but makes the image acquisition more susceptible to motion from surrounding tissues.

Conventional Doppler may be hindered by the tortuous nature of veins, preventing a reproducible placement of the gates on the image, which is not an issue with power Doppler. Vessels with very slow blood flow, as might be expected in situations where OVs are enlarged, are at the limit of the sensitivity of colour Doppler. Power Doppler can provide excellent resolution of the vein wall, and can delineate tortuous veins, but it can also fail to distinguish venous from arterial flow. There are no comparative data for the various Doppler methods and so this review cannot make any conclusions regarding the most appropriate method, but this may be irrelevant as technology continues to improve the discriminatory power and resolution of images.

Computed tomography and magnetic resonance imaging

Computed tomography (CT) and magnetic resonance imaging (MRI) are non-invasive imaging techniques. Both methods involve the body passing through an array of detectors, producing a contiguous set of cross-sectional images which can be constructed into three-dimensional images. CT involves the use of a contrast agent, which is potentially nephrotoxic, and X-rays, while MRI deploys strong magnetic forces to generate images from the resonance of water molecules. The contrast of the image can be manipulated at the processing stage, avoiding the use of a contrast agent (usually gadolinium), although the latter can be used in magnetic resonance (MR) venography by dynamic subtraction of the native and contrast-enhanced images. MR venography can produces images of a quality comparable with conventional venography. Three-dimensional T1-weighted gradient echo sequences, which can be imaged in a single breath hold, in which the varices appear hyperintense, are also an effective way of demonstrating pelvic varices with MR imaging.

The disadvantage of both CT and MRI are that they are conventionally performed with the patient in the supine position, and as with laparoscopy, pelvic varices may not be as prominent. Artefacts from metallic coils after embolisation may also limit the use of MRI for follow-up imaging, but it is likely to become the initial non-invasive investigation of choice for the diagnosis of PCS.

Hormonal treatment

There have been two small randomised controlled trials (RCTs) of MPA for women with diagnosis of PCS, with pain scores as an outcome. One showed that MPA showed a significant reduction in pain scores compared with placebo (50% reduction of pain from baseline in 73% vs. 33%, n = 104; p < 0.0001) at 4 months. At 9 months, this effect was sustained only with the support of psychotherapy. 25 In the second RCT, MPA was compared against the GnRH agonist goserelin acetate over 6 months. At 1 year after treatment, goserelin remained superior to MPA in terms of pelvic venographic improvement, improvement of sexual functioning and reduction of anxiety and depressive states as subjective measures; goserelin acetate achieved a statistically significant advantage (p = 0.0001) compared with MPA. 10 Another very small study found a benefit from a subdermal progestogen,26 but with side effects such as weight gain, acne and abnormal menstrual bleeding, which may lead to discontinuation. The orally vasoactive lofexidine hydrochloride, which was hypothesised also to impact on pain by reducing vasospasms associated with pelvic venous congestion, failed to demonstrate a benefit in a small placebo-controlled trial. 29

Hysterectomy

A case series of 36 women with PCS undergoing bilateral oophorectomy and hysterectomy reported that two-thirds had their pain eliminated while the remainder had some residual pain at 1 year postoperatively, but in only one woman was the pain affecting her daily life. 30 Conversely, Chung and Hun did not find a statistically significant reduction in pain in women who had either a bilateral or unilateral oophorectomy with a hysterectomy. 31 The failure of hysterectomy to provide adequate relief may be due to numerous anastomoses between the ovarian and uterine veins, which, if they are inadequately resected during hysterectomy, may leave residual varices. Given the trend in other benign conditions, for example heavy menstrual bleeding,32 to avoid hysterectomy when more conservative treatments are available, hysterectomy should not been seen as a first-line treatment.

Pelvic vein ligation

Ovarian vein ligation has been performed for PVI since the late 1970s, as either a unilateral or a bilateral procedure without hysterectomy. Improvements in pelvic pain were generally seen; for example, in a 2003 study of 23 women undergoing laparoscopic ligation, 78% were pain-free at 6 months postoperatively and showed no menopausal symptoms. 33 There are no randomised trials, however, and so a relative treatment effect cannot be defined. Endovascular techniques have supplanted ligation, as they are capable of being performed under conscious sedation, are quicker and have a shorter recovery period.

Pelvic vein embolisation

The embolisation of pelvic veins is via percutaneous insertion of metal coils or sclerosants such as glue or sodium tetradecyl sulphate. The aim is to permanently occlude the outflow veins mechanically and by thrombosis, ultimately causing sclerosis of the vessel. This procedure is available worldwide and the evidence for its effectiveness will be reviewed in Chapter 6.

Separation of clinical and radiological features

As with many conditions, PCS can be defined on the basis of patient-reported symptoms, clinical observations and examination and by radiological methods. The framework for assessing the definitions of PCS will reflect this, classifying the reported criteria into pain symptoms, clinical observations, pelvic venous anatomy and PVI, the last two domains being derived from ultrasound or venographic imaging. Any criteria that preclude a definition of PCS will also be noted.

Search strategy and selection criteria

Anticipating a huge number of small descriptive studies on the radiological features of PVI and of potential treatments, we restricted our review of definitions to those studies chosen for inclusion in our reviews of non-invasive imaging technologies and of embolisation of pelvic veins in Chapters 5 and 6, respectively. The selection criteria are not, therefore, specific to this review; we include the studies relevant to the key questions regarding PCS and the higher-quality studies of treatment.

Data extraction

From each included study, the criteria used to define eligibility for the study, or for treatment by embolisation, were extracted verbatim by one reviewer. The definitions were sought from the methods section of each paper, as these would be the criteria applied to the study population. Any distinction between the population selected for radiological investigation and the definition of PCS or PVI was noted. The extracted text was checked by a second reviewer and summarised to attempt a consistency in phraseology and format. The first reviewer confirmed agreement with the summary and the definitions were tabulated.

Analysis

We divided selection criteria into clinical symptoms, ultrasonographic definitions, venographic definitions and exclusion criteria. Clinical symptoms were subdivided into pain symptoms and clinical observations from examination of the patient. Radiological definitions were subdivided into anatomical criteria and descriptions of reflux. Anatomical criteria included the location, diameter and/or appearance of the veins, obtained from either ultrasound or venography. Descriptions of blood flow were derived from venography or ultrasound with Doppler. Where there were explicit thresholds for inclusion in the study, for example the minimum number of all criteria needed to be present, these were captured; otherwise it was assumed that all reported criteria were required.

We aimed either to identify corresponding or contradictory signs, symptoms or observations, or to identify criteria unique to single studies. Criteria were considered to be contradictory when the item was a reason for inclusion in one article and a reason for exclusion in another. The intention was to define a consensus group of criteria or definitions across the studies, which was arbitrarily defined at 75%: if more than 75% of studies that had a definition for PCS clearly reported any criterion, that criterion would be considered important.

Studies reviewed

The papers considered were those already selected for the reviews described in Chapters 5 and 6.

Eleven were included from the review of the diagnostic accuracy of non-invasive technique23,36–45 and 21 were included from the review of embolisation;15,31,46–64 one study was included in both reviews. 65

A summary of the inclusion criteria from the selected studies is given in Table 1.

Pain symptoms

Of the 33 studies,15,23,31,36–65 13 did not specifically define the patient population as having pelvic pain. 23,36,37,43,45–48,53–55,61,62 These studies invariably discussed definitions of PCS in the introduction section of the report, but it was not clearly stated which pain symptoms defined eligibility for the study, or else the report described the prevalence of pain symptoms in terms of percentage of patients reporting each symptom but did not indicate whether or not any were essential. In 5 of these 12 studies,38,39,46–48 the study population were women referred for screening or treatment of varicose veins of the leg.

A further 12 studies described their populations as women with chronic pelvic or lower abdominal pain with no further elaboration of the nature of the symptoms required for inclusion15,31,38,40–42,56–59,63,64 and only three specifying 6 months as the minimum duration of symptoms. 40,44,50 Six studies did describe particular pain symptoms; 38,39,50–52,65 one study from 1999 selected for the diagnostic review and the remainder describing the effectiveness of embolisation, all subsequent to 2007. Of the studies that defined symptoms, six described dyspareunia or post-coital ache as a required symptom,38,39,49–51,66 and four listed dysmenorrhoea or pain specifically just before or during menstruation. 49–51,65

Of the other symptoms occasionally highlighted, only two50,52 mentioned the exacerbation of pain after prolonged periods of standing as a defining symptom and only Carrion Otero et al. 39 included a feeling of heaviness in the pelvis or legs in the symptomology. Pain that tracked along the site of varicose veins, particularly during menstruation, was reported as defining symptoms in two studies,51,65 but was not reported in the presenting characteristics in any other studies. Only two studies required their study population to have tried, and been unresponsive to, medical treatment and to have persistent severe symptoms following laparoscopy. 30,52

Clinical observations

Examination revealing superficial varices in the pelvic region formed part of the definition of PCS in five studies. 38,39,46,53,65 Atypical varices in the thigh, vulval or buttock area were the most frequently described criteria,46,49,53,65 with no further details given regarding the size or the degree of discomfort. Three studies also accepted pelvic varices in previous pregnancies in their inclusion criteria. 38,39,49 Only one study required ovarian point tenderness to be identified in its population. 40

Venous anatomy

Ultrasonographic or venographic description of pelvic venous anatomy was a feature of the majority of case definitions in the studies reviewed, although this was not consistently defined and the terms ‘pelvic vein’ and ‘ovarian vein’ were used interchangeably. Ten studies simply stated that their populations had pelvic varices, sometimes described as dilated or voluminous. 38,41,47,50,52–56,65 More studies reported an OV diameter, above which the vein was considered to be dilated, but this threshold varied considerably, from > 4.5 mm67 to > 10 mm,57,58 with > 5 mm being the most frequently cited. 40,42,43,59,60 Halligan et al. 23 grouped patients into three ranges of pelvic vein diameter without specifying a threshold. Two studies reported an ultrasound protocol that included assessment in both a supine and a tilted or upright position, with a consequent change in the diameter as a defining criterion. 42,54

Venous blood flow measurement

Assessment of the flow of blood through the pelvic venous system was undertaken by Doppler ultrasound, time-resolved MR venography and conventional venography, although it was frequently unclear which methodology was used for each criterion. There are many ways to describe the observations of the venous blood flow, converging on general terms such as congestion, insufficiency and reflux. A non-descript term of congestion was used in nine studies,41,47,50,52–56,65 sometimes with qualifying terms such as stasis or distension, and categories of severe or moderate, but without any thresholds for such grouping. Descriptions of venous reflux were marginally better, with bidirectional or retrograde flow being indicative of reflux. The location of the reflux was precisely specified in some studies38,51,65 or else related reflux to changes in the Doppler signal observed during a Valsalva manoeuvre. 37,47,53 Filling of contralateral veins or filling described as ‘across the midline’ were indicative of pelvic vein insufficiency in nine studies,31,40,43,44,49,51,58,60,61 while six referred to filling of vulvar or thigh varices as a sign of PVI. 31,40,49,51,58,60 Velocity measurements were also a criterion, but, again, there were various criteria for abnormality, including a refilling time of < 20 seconds,39 a velocity of < 3cm per second60 or apparent stasis. 59 Only Carrion Otero et al. described the absence of venous valves as indicative of PVI. 39

Composite criteria

Several studies attempted to rationalise the various different signs, symptoms and observations by defining PCS as the presence of a certain number out of all those that the study stated as inclusion or diagnostic criteria. For example, Carrion Otero et al. 39 required at least four out of the six clinical and radiological criteria sought in the study. Creton et al. 65 required all venographic criteria, following clinical screening, in order to proceed to embolisation. A case–control study combined three criteria, each with a scale of severity from 1 to 3, to give a score, but did not compute a threshold on which to define a case of PCS. 23 Chung and Huh31 deployed a similar scoring system with four criteria, but with a threshold of ≥ 5 out of 12 as indicative of PCS. These scoring systems were similar to that first proposed by Beard et al. ,16 which was cited in one other study. 45 Scultetus et al. 53 and Asciutto et al. 49 describe four groups: group 1 with only vulval varices without PVI, group 2 with isolated incompetence of the IIV, group 3 with dilatation of the OV (although without stating what diameter constitutes abnormal) and significant insufficiency, and group 4 being those with the Nutcracker syndrome.

Exclusion criteria

Few studies explicitly stated criteria that precluded a diagnosis of PCS, or excluded women from the population. Creton et al. 65 stated that women with Nutcracker syndrome or who had varicose veins originating principally from connections with the femoral vein or who had low pain scores were excluded. Exclusion of women whose pain could be attributable to other pathological causes, such as endometriosis, was stated in five studies. 31,42,52,54,60

Summary of main findings

There was no single, clearly defined criterion for a diagnosis that was reported in the all of studies included in the review or even cited by the arbitrarily pre-defined minimum of 75%. The majority cited pelvic pain, dilated OVs and venous reflux or congestion as principal features of PCS, but many did not give thresholds or further clarification, or, where they did, these were heterogeneous. There is a need for a globally accepted diagnostic standard for PCS, which would help to standardise clinical evaluation and facilitate comparisons of outcomes of treatment effectiveness.

Strengths and limitations of the review

The review systematically extracted diagnostic criteria from relevant papers, categorised them into predetermined groups and attempted to look for consistency. A strict policy of extracting the definition from the methods section of each report was used, to extract information pertaining to the population under study. This may have meant that inclusion criteria that could be inferred from the introductory preamble were ignored and hence the frequency was underestimated. Furthermore, the studies reviewed were a subset of all those available, selected on the basis that they had been included in two other systematic reviews. In the process of screening citations for the other reviews, the authors did not locate any publication proposing a consensus diagnostic standard. It is likely that the studies here are representative and that the identification of themes has reached data saturation, to adopt quantitative research terminology.

Conclusion

A handful of studies attempted to use or create a scoring system, while two authors have proposed division of PVI into four categories, the most common of these being OV reflux. 49,53 An algorithm to diagnose and subcategorise PCS, or even to rank the severity of the condition, would be clinically useful and would help resolve inconsistencies between study reports. Such work has been undertaken in other pain-related conditions73 to determine the best combination of variables for the derivation of data-driven algorithms with optimum sensitivity and specificity relative to the reference diagnosis. With PCS, there is no independent gold standard for a diagnosis other than venography, as all current definitions include both clinical and radiological observations. A consensus diagnosis involving venography cannot be derived without incorporation bias (the inclusion of the test of interest in the reference diagnosis).

The synthesis of current literature is hampered by the absence of clear, consistent diagnostic criteria. Thus, the first aim should be to achieve consistency on the descriptions of symptoms and, from there, document comprehensive cohorts of women presenting with CPP using these definitions, and also those pathological conditions that would rule out a diagnosis of PCS. A consensus definition would improve the current situation and improve multidisciplinary working between gynaecology, where women typically present, and interventional radiology.

We propose a potential scoring template that may be adopted for reporting standards for clinical research studies evaluating the diagnostic tests for PCS and the effectiveness of treatments for PVI (Table 2). This would facilitate consistency to allow future comparison between the results of different studies, and would improve the overall quality of clinical research in this field to the standards found elsewhere in research on venous disease. 74 The proposed thresholds for considering PCS unlikely, moderately likely or highly likely are not statistically derived and represent the opinion of the authors, but they also reflect the prevailing consensus of relative importance of symptoms and observations seen in this review.

Further work is essential to refine and validate the proposed scoring system. Ideally, the diagnostic performance of each individual criterion would be evaluated for its ability to discriminate PCS from pain-free controls, and those parameters with a statistically significant difference would be incorporated into a logistic regression model. The regression coefficients of the best-fit model could be used to weight scores for each criterion. The individual and total scores would then be mapped onto a receiver operating curve to determine the cut-off value that provides the optimum clinical performance with respect to sensitivity and specificity. Finally, the scoring system would be validated in a prospective study of women with CPP to determine its accuracy. This strategy would have to accept the issue of incorporation bias and also be contingent on the use of modern ultrasound methods. This would then provide a practical, valid tool for clinical use that would assist in the identification of women for invasive confirmatory venography and, potentially, treatment. In the meantime, the proposed template may be considered useful.

Strength of association

The stronger the association between risk factor (incompetent pelvic veins) and outcome (CPP), the more likely it is to be causal. A case–control study of women with and without CPP who have undergone a valid method of pelvic vein imaging, with dilatation and reflux defined by acceptable criteria, would provide data on the odds of CPP being associated with and without PVI. A highly significant odds ratio would provide strong evidence of association.

Consistency of findings

If an association is replicable in different populations, whether defined by ethnicity, parity or country of study, this provides reassurance that any observed association is not confounded by the study population.

Specificity of the association

If incompetent veins were frequently observed alongside another known or presumed cause of pelvic pain symptoms, for example endometriosis, it would be difficult to establish whether or not each risk factor independently caused CPP, whether or not there was synergy or if one was a confounding factor of the other. Studies of women with CPP subdivided into groups according to whether or not pathology was observed at laparoscopy would assist with this distinction.

Temporal sequence of the association

To suggest a causal link, the emergence of incompetent veins should occur before the onset of pelvic pain. Strong evidence for temporality would require the screening of asymptomatic women for pelvic vein anomalies and following their life course to identify those who develop CPP, whereas pelvic venography is usually instigated as a result of CPP.

Biological gradient

An observation of greater pain with a higher degree of dilatation or severity of reflux, or number of vessels involved, may imply causation. Yet pain is a highly subjective outcome and, if viewed from a biopsychosocial perspective, organic pathology, anxiety trait and coping skills after being given the diagnosis could all be expected to contribute to a woman’s experience of pain, which may then modify any relationship between anatomy and symptom severity.

Biological plausibility

A potential physiological mechanism that links pelvic vein dilatation or reflux to the experience of pelvic pain would add weight to the evidence for causation. Pelvic varices are seen as an engorgement of blood, pooling in the pelvic veins, which is consistent with the sensation of a heavy and dragging pain described by women diagnosed with PCS. The pain often improves when a supine position is adopted, as the venous dilatation resolves with draining of the venous system.

Coherence

The analogy with saphenous varicose veins, which can cause feelings of heaviness, aches and discomfort in the leg, supports the hypothesis that incompetent pelvic veins can cause CPP.

Experiment

A convincing piece of evidence would be if the removal or improvement of reflux or dilatation in the pelvic veins resulted in a sustained decrease in the perceived CPP. Interventional radiology techniques such as embolisation can selectively block pelvic veins. There are considerable data on the reduction of pain following embolisation, but the quality of the evidence is doubtful. The efficacy of these techniques will be considered in Chapter 6.

Association of pelvic pain and pelvic vein incompetence

All reviews of the association between PVI and CPP, and the diagnosis PCS, refer back to the seminal paper of Beard et al. 16 His team performed transfundal venography on 81 women, although 18 were excluded owing to a failure to produce adequate images. The women were classified into three groups: 45 women with CPP and no apparent pathology at laparoscopy (the presumed PCS group), 10 with CPP and endometriosis, adhesions or pelvic inflammatory disease, and eight with no pain and no pathology observed during laparoscopic sterilisation as controls. Owing to the nature by which the controls were chosen, they were older on average and all were parous. Venographic parameters reported were the maximal OV diameter, contrast clearance time and congestion, which combined variation in the OV diameter and the extent to which the veins were tortuous. A ranking of 1 to 3 was assigned to different grades of severity within each variable, allowing a score to be derived.

A statistically higher OV diameter was observed in the idiopathic CPP group than in the other two groups [mean difference idiopathic CPP vs. control 3.48 mm (p < 0.005); mean difference idiopathic CPP vs. pain and pathology group 3.13 mm (p < 0.004)], and also slower clearance, with contrast still visible after 40 seconds in 26 out of 45 women (58%). Minimal tortuosity was seen in two of the eight control-group women and one woman with pathology, compared with 38 (84%) women with idiopathic CPP exhibiting moderate or extensive congestion. A venogram score of ≥ 5 gave a sensitivity of 91% (95% CI 79% to 98%) and specificity of 75% (95% CI 35% to 97%) for idiopathic CPP compared with the controls and the same sensitivity and a specificity of 100% compared with those with CPP and pathology.

This team followed up their work with another study of 68 consecutive women with a history of CPP and 36 pain-free controls undergoing sterilisation. 6 All underwent laparoscopy and most had a venogram, with 25 having identifiable pathology, of whom 22 had normal venograms. Of the 43 women with no pathology, 35 had evidence of PVI, giving a prevalence of 52% among women in pelvic pain clinics, and statistically significant negative association between pathology and PVI, although three women did exhibit both. The team went on to evaluate the relationship between particular symptoms and PVI, with post-coital ache [23/35 with evidence of PVI vs. 3/36 in control group (p < 0.0001) and vs. 2/22 with pathology and no PVI (p < 0.001)] and ovarian point tenderness [22/35 with evidence of PVI vs. 5/36 in control group (p < 0.0001) and vs. 4/22 with pathology and no PVI (p = 0.01)] exhibiting the highest discriminatory ability between those with PVI, those with pathology and those without either. The authors attribute compression of the OV and back pressure to this observation and also noted that this sign occurred on the side that the women reported was painful.

Conflicting evidence of the association of PVI and CPP came from a study of women recruited from a pelvic pain clinic, with suspicion of PVI, who underwent transvaginal Doppler ultrasound and transfundal venography. 23 Venographic evidence of PVI, defined as a score of ≥ 6 out of 9 on the PVI scoring system,16 was confirmed in 36 women. Ultrasound assessment of PVI in these women was compared with that of 19 asymptomatic women, the majority of whom underwent the ultrasound to confirm placement of an intrauterine contraceptive device but had no other organic pathology likely to cause pelvic pain. The modified PVI scores from ultrasound were unable to discriminate the venography-confirmed PVI cases from the controls (mean modified PVI ultrasound score 5.1 in PVI cases vs. 4.5 in controls; p = 0.182). The authors considered whether or not this was due to the Doppler ultrasound being insensitive to low flow rates or distorted by tortuosity, or by its use on patients in the supine position. The PVI cohort may have been defined by a threshold on the scoring system that was too low, or that contained women who had other causes of pain or were otherwise different as a result of their lower age. Furthermore, there were statistically significant differences in the uterine volume (mean difference 26.9 cm³; p = 0.008) and endometrial thickness (mean difference 4.5 mm; p < 0.001), with a small volume seen in women with PCS and a thinner endometrium, although this may be attributable to the high rate of intrauterine contraceptive device use in the control group. As with other studies,12 polycystic ovaries are more frequently observed in women with PCS (total follicular area 0.4 cm² larger in PCS group; p = 0.01).

In a Korean study, 32 women with clinical evidence of PCS, no other pathology and CPP, first had pelvic varices identified by ultrasound, with PVI confirmed by selective OV venography using > 5 mm OV diameter and various reflux features as criteria. 40 The control group, who were of a similar age, were asymptomatic and had no other pathology. The positive predictive value was maximal at 83% with a 6 mm OV diameter threshold on transabdominal TAUS, and a statistically significant increase in mean left OV diameter in those with CPP compared with controls [7.9 mm (SD 2.3 mm) vs. 4.9 mm (SD 1.5 mm) respectively; p < 0.001] was observed using this technique. By transvaginal Doppler ultrasound, 100% of the CPP group and 17% of controls had pelvic varices. There was a statistically significant difference in the mean bilateral OV diameters (e.g. a difference of 2.6 mm in the left OV; p < 0.0001) but not in the uterine volume. Polycystic changes in the ovary were also more frequent in the PCS group (41% vs. 11%; chi-squared test p < 0.006).

Seventeen women, out of 1800 who had had pelvic CT scans and no other abdominal or pelvic pathology, were described as having PVI. 78 Twelve of these women were reported to have CPP.

A large series of 725 women undergoing TAUS and 402 having a pelvic CT were assessed for venous insufficiency in the lower limbs and also questioned regarding CPP. 77 Using a threshold diameter in the OV of 5 mm, 21 women had dilated veins by ultrasound and 35 had dilated veins by CT. They found a significant relationship with CPP, with 51 of 56 (91%) of those with dilated pelvic veins reporting pain, compared with 8% of the non-dilated group (chi-squared test p < 0.0001).

A summary of these five studies is provided in Table 4. Owing to the considerable heterogeneity, the odds ratios were not pooled in a meta-analysis.

A further insight into pelvic anatomy came from a case–control study involving 25 women with a history of CPP with no obvious cause and an equal number of comparable, pain-free women undergoing a routine gynaecological check-up. 79 All women had transvaginal Doppler ultrasound that measured the resistance and pulsitility indices of the uterine artery instead of assessing venous flow. The authors found that indices were significantly lower in women with CPP than in asymptomatic controls, in contradiction to results from women with primary dysmenorrhoea,81 where these indices were lower in the pain-free controls. Again, there was no difference in uterine volume but the ovaries were not assessed. Without assessment of the corresponding venous system, the significance of these findings is unclear.

Prevalence of pelvic vein incompetence in asymptomatic populations

Populations of women undergoing detailed pelvic imaging for other reasons provide data on the incidence of PVI in asymptomatic populations. A retrospective study of the preoperative helical CT scans of female kidney donors provides data on the incidence of OV varices and incompetence in an asymptomatic population. 82 An OV diameter of > 7 mm was considered to be dilated, whilst reflux was complete opacification during the arterial phase of CT angiography; this latter observation defined incompetence. Medical note review implied that no women reported pelvic or abdominal pain, although 4 out of 34 had small fibroids. The prevalence of PVI was 47% (16 of 34 women), all in the left side, six with bilateral incompetence. The mean diameters were 9.1 mm for the left OV and 8.8 mm for the right OV. A statistically significant association of PVI with parity was observed, with 15 out of 16 with PVI compared with nine of 18 without being parous (chi-squared test p = 0.005).

A study of the preoperative angiography results of female kidney donors also provides data on PVI prevalence. 83 The authors used the venous phase of the angiogram to evaluate the OV, with a threshold of 8 mm to define a varice. Women with this diagnosis were asked to complete a questionnaire about CPP before and 6 months after they underwent left nephrectomy. Of 273 women whose aortograms were reviewed, 27 (9.9%) were found to have PVI in the left OV. Within this group, 22 completed the questionnaire and 13 (59%) reported pelvic pain before the surgery. The authors also remarked that nephrectomy requires OV ligation, used as a treatment for PCS, and noted that seven gained complete relief (54%) and three gained partial relief (23%) of pelvic pain.

Another retrospective review of pelvic MR angiograms of 22 female kidney donors found that eight had indirect evidence of reflux in the left OV, with a mean diameter of 6.4 mm (SD 0.8 mm), 13 had no reflux and 1 OV could not be located, giving a prevalence of 38%. 70 From this cohort, 19 had an adequate clinical history, confirming the absence of CPP in all.

In a large series of 324 women who had undergone multiplanar CT for reasons other than CPP, right OV variants that drained directly into the right renal vein were identified in 32 (9.9%). 84 Of these women, three exhibited left OV reflux and one exhibited right OV reflux with predominant right pelvic varices, defined as veins of > 5 mm. Dilated pelvic veins were identified in 59 of the 324 women, and reflux in 57 of these 59 woman, of which all but one were predominantly in the left OV. There was no statistical difference for the presence of pelvic varices between women with and women without a right OV drainage variant (p = 0.3). There was no significant difference in the diameter in either OV between women with and women without a right OV variant, but there was a significant difference between the parous and nulliparous women. OV reflux was identified in 17.6% of this population. These authors noted that two women did not have left OV reflux but still had pelvic varices, but they conclude that this may be due to the limitations of CT imaging, which is performed while the patient is in the supine position. This study may provide indirect evidence that drainage from the OV directly into the renal vein might not be a primary factor in the development of pelvic varices, as junctional variations of the right OV did not seem to be associated with reflux and pelvic varices.

Relationship with varicose veins of the lower limbs

Pelvic varices are often incidentally observed during investigation of varicose veins of the lower limbs, using colour Doppler ultrasound, the standard method for assessing superficial venous insufficiency. In a large cohort of women undergoing lower-limb assessment and classified using the Clinical, Aetiological, Anatomical and Pathophysiological (CEAP) criteria85 as being between zero and five, without a history of deep-vein thrombosis, 249 had TVUS with Doppler and 59 had selective venography. 37 The prevalence of reflux of pelvic origin into the lower limbs was 124 out of 1020 (12.2%), roughly equally distributed between bilateral and unilateral. In a subsample of this cohort, 150 out of 249 (60.2%) women had pelvic varices seen by TVUS with Doppler, with a mean diameter of these veins of 8.5 mm (SD 1.7 mm). The positive predictive value of reflux of pelvic origin in the lower limb for observation of a pelvic varice was 49%, and negative predictive value was 94%. Assessment of the lower limbs, therefore, cannot be the sole test for PVI, and so the authors suggest that pelvic pain be considered before further TVUS investigation. However, whenever a negative result from colour Doppler ultrasound of the lower limbs is observed, further investigation of pelvic veins is not necessary. A statistical significant relationship between lower-limb reflux of pelvic origin and recurrent varicose veins was observed, suggesting an important and unrecognised cause of recurrence. 37

A similar study of 100 women, all reporting CPP of unknown origin, investigated the relationship pelvic varices and previously undiagnosed lower-limb venous insufficiency. 81 Using a threshold of 5 mm OV diameter, the prevalence of varices was 30%, with a positive predictive value of lower-limb insufficiency for pelvic varices of 77% and negative predicative power of 87%. An additional observation was that there was no difference in the incidence of ovarian cysts between those with and those without OV varices.

The series of 1029 women described above were also assessed for venous insufficiency in the lower limbs. 77 Using a threshold diameter in the OV of 5 mm, 56 women had dilated veins and lower-limb venous insufficiency was found in 44 of these.

Summary of key findings

Overall, there is a lack of well-designed case–control studies to delineate the relationship between CPP, dilated pelvic veins and PVI. In the five case–control studies with appropriate data amongst those assessing the association between pelvic pain and PVI, the associations were generally fairly similar,6,16,23,40,77 with three studies showing statistically significant associations (odds ratios of between 31 and 117). 6,16,77 The two smallest studies failed to reach statistical significance in the odds of association, perhaps because they were too small for a difference to be detected. The proportion of women found to have PVI who reported CPP ranged considerably, from 39% to 91%. Polycystic ovaries were observed more frequently in the group with CPP and PVI in two studies. The prevalence of PVI ranged considerably, although the majority of women with PVI had CPP. Conversely, in the four studies of asymptomatic women, no more than half had PVI, although again the prevalence ranged widely. Where lower-limb venous insufficiency was seen, between 60% and 77% of women also had pelvic varices.

Strengths and limitations of review

We could be criticised for not conducting a search specific to this review question, potentially missing important citations. In screening several thousand citations for the reviews of imaging methods and the effectiveness of embolisation, we located all of the papers included here. Moreover, in the process of reviewing the shortlisted studies for these other reviews, we scanned the reference list for additional relevant papers and did not locate any additional contemporary papers. Some papers refer to classical studies such as those by Taylor and Hobbs,8,11 but these predate the advent of diagnostic laparoscopy or deployed insensitive imaging techniques. We are therefore confident that we have located all relevant case–control studies, although we may have missed some studies which focused primarily on lower-limb insufficiency, with the coexistence of PVI a secondary consideration.

The six most pertinent studies were unfortunately of mixed methodological quality, drew on clinically disparate populations and defined PVI inconsistently, making it inappropriate to pool the results. The first study of Beard et al. 16 used a control group who were older and of higher parity, which may have confounded the groups, while other studies retrospectively identified the pain status of the participants after venography, with considerable potential for the underidentification of pain. The selection of controls for this type of study can undermine the conclusions drawn if they are either too disparate or overmatched. Ideally, there should be considerably more controls than cases to improve the power of the analysis, whereas the reverse was true in the studies by Beard et al. 16

Interpretation

In reviewing the data for evidence of a causal association between PVI and CPP, we should refer back to the criteria of strength, consistency, specificity and a biological gradient, which can be addressed by the studies located in our review.

The studies of the association between PVI and CPP are varied, making us reticent to conduct meta-analysis and denying us the opportunity to derive a pooled estimate of the odds of association between PVI and CPP. The study by Park et al. 40 perhaps represents the most robust data, with standardised contemporary imaging techniques and comparable controls in terms of age, but defines presence of reflux from ultrasound rather than venography, potentially miscategorising some of those whose blood flow appeared normal. The difference in left OV diameter was statistically significant between the groups, but unfortunately the data could not be disaggregated to construct a receiver operating curve to determine an optimal threshold, and, overall, the study had only 7% power to detect a statistically significant association given the odds ratio observed. Other studies produced highly significant odds ratios that would suggest a strong association between PVI and CPP that is possibly consistent across study settings. The ambiguity and inconsistency with which PVI is defined, as discussed in Chapter 3, may explain the differences between studies, as might the imaging method, with supine ultrasound potentially under-recognising reflux. As no study conducted an a priori sample size calculation, it is also possible that a significant relationship has been missed in some of the smaller studies.

The studies of Beard et al. 6,16 provide some reassurance that PVI is not frequently observed alongside other pathology such as endometriosis, and, therefore, the pain could possibly be attributed to the observed PVI. In their first study, the odds of association were higher when comparing the cases against the group of women with CPP and pathology than against the non-pain control group, although numbers were very small. This was replicated in their second study, which compared an idiopathic CPP group against a group who mainly had confirmed endometriosis.

There are no prospective cohorts to address the issue of temporality or studies that attempted to rank PVI severity and pain intensity. All cases were defined by a clinical history of CPP and, unlike some of the studies of embolisation, did not ask women to complete visual analogue scales to score their pain. Pain is highly subjective and individual, and so might never exhibit a proportional relationship with any ranking of PVI severity. An analogous situation occurs in endometriosis, whereby the relationship between the degree of endometriosis and the amount of pain experienced by the woman does not always seem correlated. 86

The study of potential kidney donors is a fortuitous method of establishing the prevalence of PVI in general populations, and in the left OV in particular, as laparoscopic donor nephrectomy tends to be performed on the left kidney. Prevalence of PVI in the left OV was variously reported at 10%, 17%, 18%, 38% and 47%. 70,82–84,87 The varying thresholds for defining OV dilatation may give rise to these differences, as might the imaging systems deployed, although no apparent association can be deduced here. CT may overmagnify vessel diameters by 20% yet be unable to detect small veins, and some distention of the OV may occur as a result of the increased intra-abdominal pressure caused by breath holding. Two studies both used CT and a threshold of 5 mm to define dilatation in the pelvic veins and yet prevalence of varices were 9% and 18%, respectively. 77,84 Belenky et al. 83 argued that detected of retrograde venous flow by renal artery angiography is preferable as it is physiological, whereas selective venous catheterisation may aggravate reflux. Nascimento et al. 70 remarked that the MR angiography technique they used was suboptimal and led to less indeterminate determination of blood flow, in addition to the general issue of undertaking the scans in the supine position and with breath holding.

The variation may also be explained by factors such as parity, age and ethnic group, but it illustrates that PVI is ubiquitous and relatively common. Ahlberg et al. 88 examined the venous structure in cadavers and reported the complete absence of valves in 15% of left OV and 6% of right OV, with half of women exhibiting incompetent valves.

Belenky et al. 83 provide good data on the incidence of OV varices in an unselected population, but they do not support the hypothesis that OV varices are invariably symptomatic, with 41% of patients not reporting CPP prior to their nephrectomy, suggesting that other mechanisms contribute to PCS. With retrospective recall of preoperative pain, this finding may be subject to recall bias and also precluded further examination by a gynaecologist to explore other sources of CPP. Furthermore, 4 of the 22 responders were peri- or postmenopausal, with others claiming that PCS is a condition experienced solely by women of reproductive age. 6 Finally, the study has the same limitation as that noted with CT imaging, in which the supine position and inability to perform the Valsalva manoeuvre underestimates the incidence of reflux.

There appears to be a significant relationship between the presence of pelvic varices and lower-limb venous insufficiency, suggesting that where pelvic varices are identified, the superficial veins should also be assessed with colour Doppler ultrasound, as the underdiagnosis of leg varicosities is likely. Early identification would prompt treatment, reducing the likelihood of consequent ulceration and unsightly leg veins. Evaluation of the pelvic venous system when lower-limb venous insufficiency is found does not yield very many additional observations of pelvic varices in the absence of pain.

Conclusion

There is insufficient evidence to categorically state that PVI is the cause of CPP in women with no other pathology. There are strong associations in some studies, although there is significant heterogeneity among the reviewed studies, which also have some methodological weaknesses related to the control groups. A large, well-designed case–control study, using contemporary imaging techniques and clear criteria for the ascertainment of both CPP and PVI, is required to help to remove the residual doubt about the relationship between the two phenomena.

Search strategy

A thorough and comprehensive literature search was developed. This was used in the following widely recognised databases: Cumulative Index to Nursing and Allied Health Literature (CINAHL), The Cochrane Library, Database of Abstracts of Reviews of Effects (DARE), EMBASE (Excerpta Medica Database), Latin American and Caribbean Health Sciences Information System (LILACS), MEDLINE (Medical Literature Analysis and Retrieval System Online), Medion, Scientific Electronic Library Online (SciELO) and Web of Science. International databases including African Index Medicus (AIM), Index Medicus for the Eastern Mediterranean Region (IMEMR), Index Medicus for the South-East Asian Region (IMSEAR), Pan American Health Organisation (PAHO), Population Information Online (Popline) and Western Pacific Region Index Medicus (WPRIM) hosted by the World Health Organization were also searched. Searches were from database inception to March 2014.

Our search term combinations consisted of medical subject heading subheadings, text words and word variations. For our population, the terms ‘pelvic congestion syndrome’, ‘chronic pelvic pain’, ‘dyspareunia’ and ‘dysmenorrhoea’ were then combined with terms for imaging tests including ‘ultrasound’, ‘magnetic resonance imaging’, ‘computer tomography’, ‘angiography’, ‘phlebography’ and ‘venography’. Wildcard characters were used to capture alternative spellings and stems of words. This search strategy was modified to suit to each database being searched. The search was restricted to ‘humans’ and ‘females’. Bibliographies of all relevant primary articles and reviews were also hand-searched to identify any articles missed by the electronic searches. A comprehensive database was constructed using Reference Manager 12.0 (Thomson ResearchSoft, San Francisco, CA, USA) to store all identified references. No language restrictions were applied in the searching process.

Study selection and data extraction

Studies eligible for inclusion in the review were selected in a two-step process. First, citations identified by the electronic database searches were screened, predominantly by a single reviewer, based on their titles and abstracts, for relevance to pelvic pain. Full manuscripts were obtained for those citations that met or potentially met the predetermined inclusion criteria. Two reviewers then independently inspected the manuscripts to confirm that they fulfilled the following criteria:

1. Population – women with a clinical diagnosis of CPP. No restriction was placed on any previous treatment, age of the participants, duration of symptoms, comorbidity (including the copresence of endometriosis or other gynaecological cause) or severity of the complaint in selection of the studies.

2. Index test – any non-invasive imaging test, including TAUS or TVUS with or without Doppler (duplex, colour, power), MRI, MR venography, CT and CT venography. Phlebography was accepted as an equivalent term, while descriptions of angiography were reviewed to see if a venogram was actually performed. Where conventional CT or MR scans were undertaken before CT or MR venography, respectively the latter was considered to be the index test.

3. Reference test – confirmatory diagnosis by fluoroscopic venography. Venography could be transfundal (injection into the myometrium of the uterus) or transcatheter (via the femoral, jugular or another vein). Phlebography was accepted as an equivalent term, while descriptions of angiography were reviewed to see if a venogram was actually performed.

4. Outcome – diagnosis of PVI, by whatever method or criteria.

5. Study design – studies in which index test and reference standard were performed in the same individuals. We anticipated that studies may not have been designed as classical test accuracy studies, and may use the index test as a method of screening prior to venography and potentially embolisation. These were included in reviews where data were reported on the index test findings.

Studies were excluded if the reference standard was not performed alongside (or within 1 year of) the index test. Similarly, studies were also excluded if only the reference standard was undertaken or where there was insufficient description of the findings of the index test to allow attempts at data extraction for a 2 × 2 table. Retrospective studies, where the population was those undergoing embolisation and where it could not be determined how many patients had initially had the index test, were also excluded.

Two reviewers independently assessed the full-text papers to determine if they met the above criteria. Any disagreements surrounding the eligibility of a paper were solved through consensus. Data were extracted on study characteristics and methods, with any limitations described by the authors noted. Data were collated on study characteristics, including methods of recruitment, patient characteristics, outcomes and results. If a study undertook more than one imaging test, the data were extracted separately. Results from the studies were used to generate 2 × 2 tables where possible, cross-classifying index test results and venographic diagnoses.

Methodological quality assessment

The methodological quality of all the papers fulfilling the inclusion criteria was assessed by both reviewers and consensus achieved. Quality was defined as the confidence that the study design, conduct and analysis minimised bias in the estimation of test accuracy. Quality was assessed using the Quality Assessment of Diagnostic Accuracy Study (QUADAS) tool. 89 The original publication of this checklist identified 14 items for assessment; 11 of these are included in the Cochrane version of the tool and it is this version that was used for this review. The three excluded QUADAS items relate to the quality of reporting rather than to methodology. We further chose to not include the question regarding whether or not the same clinical data would be available when test results were being interpreted, as would be available when the test is used in practice, as we felt that the nature of the tests being evaluated and the clinical context made this redundant. Therefore, our quality tool consisted of 10 quality questions, which are described in Table 5.

Data synthesis

Data on the number of women with CPP found to have or not have PVI on non-invasive imaging were used to populate 2 × 2 tables, extracted from the study reports in duplicate. Where possible, the sensitivity and specificity were calculated. The objective was to obtain pooled estimates of sensitivity and specificity where it was possible and appropriate to do so. We anticipated three types of outcome: a continuous measure of OV diameter, potentially with different studies using different threshold in their definition of excessive dilatation; a dichotomous description of either dilatation, tortuosity or reflux; or an ordinal outcome of the degree of reflux, such as slow, reversed or absent blood velocity.

A meta-analysis was considered if there were more two or more studies using the same non-invasive test and providing sufficient data to extract data into a 2 × 2 table for any one outcome. Depending on the type of outcome, the summary statistics that would be estimated would be either the pooled sensitivity and specificity for dichotomous outcomes or the expected receiver operating curve (ROC) curve for a test across many thresholds (the summary ROC curve) for continuous variables, using Review Manager 5.1 (2011, The Cochrane Collaboration, The Nordic Cochrane Centre, Copenhagen, Denmark). Ultimately, meta-analysis was not considered possible.

Study selection for the review

A total of 6016 citations were identified by the electronic searches. Of these, 12 articles (involving 1579 women) satisfied the inclusion criteria and were included in our review. Figure 1 shows the flow of studies from identification in the literature searches through to inclusion in the systematic review. A total of 6016 citations were identified from the bibliographic database searches and one citation was found through hand-searching. Of these, 5882 were excluded after screening of titles and abstracts. This left 140 citations for which full-text manuscripts were retrieved. Of these, 128 were excluded, as they were reviews, comments or case reports, they considered the anatomy of pelvic veins in relation to the veins of the lower limbs or there were no usable data. This left 12 studies which fulfilled our inclusion criteria and were included in our review.

FIGURE 1.

Study selection process for the systematic review of the accuracy of non-invasive tests for identification of PVI.

Description of study characteristics

Table 6 provides a summary of the characteristics of the studies included in the review. Sample sizes varied from 6 patients to over 1000 in studies exploring the association between lower-limb varicose veins and pelvic varicosities. Three studies purported to be assessing the accuracy of ultrasound against venography,41,42,45 with a further two being case–control studies of the association of PVI and CPP. 23,40 Two studies assessed the accuracy of MR venography against conventional venography. 38,43 Three studies principally reported on the effectiveness of embolisation but included some data on ultrasound as a screening test. 36,44,65 Two studies looked the relationship between PVI and lower-limb venous insufficiency. 37,39

Demographic data on the participants of the selected study are given in Table 7. Information was patchy, with some studies not describing their populations. The majority of women were of reproductive age, although in one study the oldest participant was 66 years old. Two of the three case–control studies had groups who were not comparable in terms of mean age. Information on parity was more scarce, with only 5 out of 11 studies reporting any data on parity. The percentage of nulliparous women included in the studies ranged from 11% to 54%.

Quality assessment

A breakdown of quality items assessed in the 12 included studies is provided in Table 8. We considered the possibility of spectrum bias in the two studies recruiting from varicose vein clinics, as patients would not normally be considered for pelvic vein assessment if they did not report CPP. Although all studies used venography as the reference standard, and this appeared to be performed consistently among all those who underwent it, there were three studies that did not adequately define PVI,36,37,41 meaning that we could not be clear about whether or not all participants were verified against the same criteria. All but two studies did not describe the interval between the reference standard and index test. The significance of this is unclear, as the natural progression of PVI is unknown. By the nature of the tests, the index and reference tests were performed and interpreted separately.

There was partial verification in five studies, and two further where it was not possible to determine whether all index tests had been verified. In these studies, only those in whom the index test indicated PVI were taken forward to undergo venography. This means that the proportion of false negatives and the sensitivity of the index test cannot be calculated. In these studies, it is implicit that the reference standard is interpreted with knowledge of the index test, and, indeed, only two studies explicitly described blinding, both of which used MR venography. There were a few instances in which a pelvic vein could not be identified by either imaging technique, but this did not necessarily mean an indeterminate response. In the total absence of a flow chart of patients’ progression through a study, it is impossible to establish whether or not any women were excluded from the analysis. As some studies were stated as retrospective, it is likely that some selection of patients occurred and the populations were not consecutive.

Accuracy review

Owing to a lack of data within the included studies, it was not possible to generate full 2 × 2 tables for the majority of studies. Therefore, results will be discussed narratively.

Standard ultrasound

The first use of ultrasound in diagnosis came from the same team that produced the seminal paper on the association between PVI and CPP. 41 In this study, 73 women with CPP underwent pelvic ultrasound scans, and pelvic venography was undertaken in 41 of these women. Venography identified varicosities in 33 of the 41 women and ultrasound detected varicosities in 31 of these 33 women, a sensitivity of 94%. It was not possible to compute the specificity. Thirty-two women had laparoscopy and not venography, while 14 women had both venography and laparoscopy. Polycystic ovaries were reported in 44 of the 73 women.

Giacchetto et al. 42 described a study of 35 women in whom there was no pathology that could explain the clinical symptomatology. All women underwent TVUS examination but the proportion in whom dilated veins was observed was not provided. The presence of pelvic varices was confirmed by left OV venography in 16 of 35 women (46%) of cases, and was bilateral in three.

Ultrasound with Doppler

In studies with a focus on embolisation, the description of the diagnostic work-up is poor, and the studies of Bachar et al. 44 and Creton et al. 65 are no exception. Only four patients underwent TAUS and venography in the former44 and while PVI was seen in the left OV in three and bilaterally in one immediately before embolisation, the ultrasound findings are not discussed, other than to say that they concurred with venography in failing to see PVI on the right side in the three patients. Creton et al. 65 focused on the anatomical distribution of the ultrasonically identified varices, with all 24 patients exhibiting perineal varices, of which one was bilateral.

In an earlier study of 30 women, 23 were confirmed by venography to have PVI, and, while it was stated that there was agreement between TAUS and venography in 22 cases (77%), these cannot be disaggregated, as the description implied that the agreement could be positive or negative. 45

Carrion Otero et al. 39 conducted one of the largest studies, recruiting women with lower-limb venous insufficiency and carefully categorising them by their ultrasonic and venographic results. Of the 233 patients, 15 had pelvic veins of < 8 mm by TVUS and were classified as normal and apparently did not undergo venography. In the remainder, five were subsequently reclassified as normal, giving a positive predictive value of 98%.

The population in the study of Halligan et al. 23 has been described in Chapter 4, but here we are not concerned with the control group. Scoring systems of 3 (normal) to 9 (severe PVI) were described for both venographic and ultrasound observations, but only those 36 with a score of six or more by the venographic score constituted the study population. Using the same threshold for the ultrasound score, a sensitivity of 51% (18 out of 35, with no explanation for the one missing patient) for ultrasound can be deduced, but specificity cannot. 23

Another study, also previously considered in Chapter 4, assessed 139 patients with clinical suspicion of PVI who were scanned by both transabdominal and transvaginal approaches. Of those, 74 were described as having pelvic varices without other pathology. Unfortunately, only 32 of these underwent venography, with no rationale as to why only a subset went forward, and so not even a positive predictive value could be computed. 40

The most comprehensive analysis of the accuracy of ultrasound came from Barros et al. 37 Pelvic varices were identified in 150 of the 249 patients by TVUS with Doppler and in 156 patients by venography. The diagnostic accuracy parameters of TVUS with Doppler were 96% sensitivity (95% CI 92% to 99%) and 100% specificity (lower 99% CI 97%), with positive and negative predictive values of 100% and 94%, respectively. 37

Magnetic resonance venography

Two studies of reasonable quality, albeit small, explored MR venography. 38,43 The first prospectively recruited women with suspicion of PVI in whom there was no infrainguinal source for lower-limb varices observed on ultrasound. 38 MR venography was undertaken in the supine position and venography in the reverse Trendelenburg position, with all images assessed by a single radiologist, blind to the other test and assessed in random order. PVI was defined separately for three anatomical locations: dilated OV of at least 1.5 times the diameter of the contralateral vein, contrast depicting the pelvic plexus or dilatation and tortuosity of the IIV, and graded into three categories of severity. 90 Images obtained by MR venography were all of satisfactory quality and gave a sensitivity of 88%, 100% and 91% of PVI in the OV, IIV and pelvic plexus, respectively, and corresponding specificity of 67%, 38% and 42%, respectively.

The second study was a retrospective review of 19 premenopausal women who had had both MR and conventional venography on the basis of clinical suspicion of PVI. 43 A clear definition for PVI was included, and prompted 16 of the patients to have embolisation following the conventional venography; the remainder did not have the procedure as they were identified as having Nutcracker syndrome. All patients had some degree of reflux, and so the accuracy parameters reflect the ability of MR venography to distinguish grade 1 reflux, which was reflux confined to the left OV and/or parauterine veins, from grade 2, which additionally included reflux in the right OV, both IIVs and varicosities of the vulva and thigh. The MR venography showed excellent agreement between the two independent observers and a sensitivity of 67–75% for MR venography to detect grade 2 reflux. As there were no patients without any reflux, probably due to the retrospective nature of the study, the specificity was unreliably estimated at 100%.

Main findings

Transcatheter selective venography performed in a tilted position remains the gold standard for the determination of PVI and is an integral part of pelvic vein embolisation, and so it remains an essential part of the diagnostic process. TVUS with Doppler has a place in the screening of women for pelvic varices, although the data on accuracy are limited. One study provided diagnostic accuracy parameters of 96% sensitivity (95% CI 92% to 99%) and 100% specificity (lower 95% CI 97%), with positive and negative predictive values of 100% and 94%, respectively. 37 Similarly, MR venography appears to have good accuracy, but data are limited to one study suggesting 88% sensitivity and 67% specificity for identifying PVI in the OV. 38 Imaging modalities that demonstrate both venous dilatation and reflux are necessary.

Strengths and limitations

The strength of this review is the extensive literature search, ensuring that the maximum number of appropriate studies were identified. A quality appraisal of the selected studies was conducted in parallel, and although there are no concerns about the verification method, using venography, it is clear that there are several issues common to most studies that undermine their estimates of accuracy. Only two studies were prospective and included consecutive patients,23,38 while in about half, not all of the patients who underwent ultrasound assessment had the findings verified by venography. It was assumed that in order to progress to treatment with embolisation, the venography would have had to confirm the presence of PVI. Invariably, because ultrasound was being used as a triage test to identify those patients who would go on to have venography and embolisation, those who did not show evidence of PVI on ultrasound never had their status verified, and hence specificity cannot be calculated. In these and other studies, another issue was that the venography was performed and analysed with knowledge of the initial test results, which could influence the interpretation. Only the study of Asciutto et al. 38 was rigorous in its methods to prevent ascertainment bias in the analysis of the images.

Interpretation

The normal appearance of the pelvic venous system is of OVs arising from the ovarian plexus, joining the inferior vena cava on the right side and the left renal vein on the opposite side. Incompetence or absence of veins in the ovarian artery can lead to retrograde flow and progressive dilatation of the vein, creating bulging and tortuous varicosities. For a diagnosis of PVI and potential treatment, the pelvic veins need to be demonstrated not only to be enlarged but also to have poor venous flow; therefore, for identification, a dynamic imaging method is required. Only two studies in our review, both conducted over 15 years ago, used conventional ultrasound. 41,42 Commentators at that time highlighted the poor sensitivity of this method,11 which is probably due to the restriction a diagnosis of venous incompetence to those with a OV diameter of > 5 mm, missing subtle signs of reflux. We did not locate any direct comparisons between abdominal and vaginal approaches for ultrasound, but intuitively the transvaginal route is likely to give higher-resolution images, particularly for women with high adiposity.

Including a Doppler study in an ultrasound assessment should add significant value to the assessment, and the results of Barros et al. 37 suggest that this is highly accurate. The ability to assess the Doppler profile during the Valsalva manoeuvre is an important feature. Performing the Valsalva manoeuvre increases intraperitoneal pressure and a variable Doppler waveform is seen. In the context of women with suspected PVI, reversal of flow and an abrupt disappearance of the flow correlate with reflux and stasis seen at venography. 40 This is a corollary of the use of the Valsalva manoeuvre in the diagnosis of scrotal varicocele in men. 91

The precise technique may be important. Conventional Doppler may be hindered by the tortuous nature of veins, preventing a reproducible placement of the gates on the image, which is not an issue with power Doppler. Vessels with very slow blood flow, as might be expected in situations with enlarged OVs, are at the limit of the sensitivity of colour Doppler. Power Doppler can provide excellent resolution of the vein wall, and can delineate tortuous veins, but it can also fail to distinguish venous from arterial flow. There are no comparative data for the various Doppler methods, and so this review cannot make any conclusions regarding the most appropriate method, but this may be irrelevant as technology continues to improve the discriminatory power and resolution of images.

Magnetic resonance venography appears to be a very accurate imaging method, and has some advantages. The isotonic nature of the MR contrast agent reduces the risk of thrombosis associated with iodinated contrast as well as reducing the inherent risk of anaphylaxis. MR venography appears to be reproducible, with moderate agreement with conventional venography for anatomical features and a slightly lower agreement for classification of incompetence. CT may also offer similar benefits, but no studies comparing this method with conventional venography were identified. MR venography is apparently highly sensitive in identifying PVI, but, as with laparoscopy, it has compromised specificity owing to the scan being performed in the supine position. The accuracy needs to be confirmed in larger cohorts of well-designed studies but is already becoming widely used.

Scoping search

A scoping search in PubMed using the term ‘pelvic congestion syndrome’ retrieved 24 uncontrolled studies involving coil embolisation of pelvic veins and seven studies involved sclerotherapy. Seven studies of OV ligation were identified in an initial scoping search. We made a decision, after registration of the review on the PROSPERO database, to exclude studies of ligation from this review, as this procedure is rarely performed in current practice now that the technically less demanding and lower-risk option of percutaneous vein embolisation is widely available.

Search strategy

A comprehensive search strategy was developed. This was used in the following bibliographic databases: Web of Knowledge, British Nursing Index, CINAHL, The Cochrane Library, DARE, EMBASE, Medion, MEDLINE and Web of Science. The foreign databases AIM, IMEMR, IMSEAR, LILACS, PAHO, Popline, SciELO and WPRIM, held on the World Health Organization portal, were also searched. All databases were searched from database inception to November 2013. Bibliographies of all relevant primary articles and reviews were hand-searched to identify articles missed by the electronic searches. No language restrictions were applied during the searching phase.

Search terms for the condition included ‘pelvic pain’, ‘pelvic congestion’, ‘pelvic or ovarian vein’, ‘incompetence or reflux’, and variations of these as keywords and text. Search terms for the intervention included ‘treatment’, ‘endovascular therapy’, ‘interventional radiology’, ’embolisation’, ‘sclerotherapy’, ‘ligation’ and ‘occlusion’, and variations of these as keywords and text. Wildcard characters were used to capture alternative spellings and stems of words.

The condition and treatments terms were each combined using the ‘or’ term to broaden the search and the two components were combined using the ‘and’ function. There was no restriction on study design in the search, based on the output of our scoping search which indicated a dearth of RCTs (see Appendix 1).

A comprehensive database was constructed using Reference Manager to store all identified references.

Study selection and data extraction

Studies were selected for inclusion in the review in a three-step process (Figure 2). First, the citations identified by the electronic literature databases searches were screened by one reviewer, who read the titles and abstracts and selected on the basis of the population and intended treatments, and who eliminated all studies that did not appear to describe embolisation or sclerotherapy of pelvic veins. A significant number of citations relating to varicose veins of the leg were anticipated owing to the necessary inclusion of ‘iliac vein’ as a search term. Any citations referring to treatment of the saphenous vein were excluded, unless the abstract referred to the concurrent treatment of pelvic veins.

FIGURE 2.

Study selection process showing the identification of eligible studies for the review of effectiveness of embolisation.

The second round of screening involved two reviewers, who selected citations that included patients described as having PCS or PVI. Full manuscripts were then retrieved of citations that met or were thought to have met the inclusion criteria. Two reviewers then independently inspected the manuscripts to confirm that the following criteria were fulfilled:

1. Population – women with a clinical diagnosis of PCS and/or radiological diagnosis of PVI, with or without CPP. No restriction was placed on any previous treatment, age of the participants, duration of symptoms, comorbidity (including the copresence of endometriosis or other gynaecological cause) or severity of the complaint in selection of the studies. No restriction was made on the method of identification of the pelvic varices to be embolised.

2. Interventions – coil embolisation or sclerotherapy of pelvic veins. No restriction was placed on the method of embolisation or sclerosant used.

3. Outcomes – the primary objective was to assess the impact of embolisation on pelvic pain, so studies reporting subjective assessment of pain or improvement in pain symptoms were included here. Secondary outcomes were radiological assessments, for example pelvic vein diameter, complication and adverse effects, patient-reported general improvement or quality of life.

4. Study design – ideally, only reports of well-designed RCTs would be included. Preliminary searching indicated that there would be few randomised or well-designed non-randomised controlled studies. Primary reports of observational studies were, therefore, included, but restricted to those in which participants were recruited prospectively. Case reports or small series with fewer than six participants were excluded. Where it was ambiguous whether the data was collected prospectively or retrospectively, a judgement was made based on the time frame of the follow-up assessments. Where participants were reassessed at particular time points, it was assumed that they were recruited into the study prospectively. If the duration of follow-up was reported as a wide time frame, it was assumed that participants had been identified retrospectively and the follow-up was a cross-sectional survey; therefore, these reports were excluded.

Both reviewers extracted data on study characteristics and methods, and noted any limitations described by the authors. Any disagreements surrounding the inclusion of a manuscript were resolved through consensus.

Methodological quality assessment

All manuscripts selected for inclusion were assessed for their methodological quality in duplicate. This was defined as confidence that the study design, conduct and analysis minimised bias in the estimation of effectiveness. For any randomised trials found, the risk of bias tool developed by the Cochrane Collaboration was used. 94 A randomised study was considered to be of high quality if it provided evidence of adequate randomisation sequence generation and allocation concealment, if blinding was used, if the issue of missing outcome data was adequately addressed and if the published paper was free of selective reporting and other biases.

Acknowledging the lack of RCTs, the quality assessment was redefined for observational case series. These are the least methodologically sound study designs and have inherent problems that can bias causal inference owing to the fundamental problem of a lack of a comparison group. There are no universally accepted quality criteria for case series, as systematic reviews usually tend to exclude studies of this design, and there are no reporting standards. The quality of case series was assessed by considering the following criteria, adapted from a published checklist,95 such as they were reported, described in Table 9.

Data synthesis

Standard meta-analysis methods were to be used to estimate the overall proportion with a symptomatic improvement following embolisation, using the proportions reported in individual studies and under fixed- and random-effects assumptions. MedCalc software (version 14.10.2, Ostend, Belgium) uses a Freeman–Tukey transformation96 to calculate the weighted summary proportion under the fixed- and random-effects model. 97 Plots were generated in MedCalc. For other outcomes, studies were extracted in duplicate, tabulated and described narratively.

Studies selected for the review

A total of 2858 citations were identified through the electronic bibliographic database searches. Of these, 2718 were excluded after the titles and abstracts were read, mainly because they referred to varicose veins of the lower limbs. This left 140 citations for which full papers were retrieved.

Nine were eliminated, as they were retrospective in design. If there was ambiguity regarding whether the study was prospective or retrospective, and it was available only as an abstract, the citation was excluded. Four studies, in non-English journals, had limited translation which indicated wide follow-up time periods, and were excluded on the assumption that they were retrospective in design.

Seven studies where pelvic vein ligation was investigated were identified. The original objective had been to include such studies in the review and search terms relating to pelvic vein ligation were retained in the search strategy to maximise the retrieval of studies of intervention. Of the seven studies, one was a case report of a single procedure, two were in Chinese, one reported on open procedures that are no longer practised and one was a study of female kidney donors and the association between pelvic varices and CPP. The remaining two studies considered more appropriate laparoscopic pelvic vein ligation, reporting on 50 patients in total, with one study making a non-randomised comparison against medical treatment.

Three studies were excluded at the final stage because they were considered retrospective, but as they reported on medium- to long-term outcomes they were considered for this outcome. 62,98,99 Two studies, with a total of 37 patients, although prospective in design, reported only on technical success and complications, and so were excluded. 100,101 This left 22 studies, reporting on 1308 patients. 15,31,46–65 A summary of the characteristics of the included studies is given in Table 10.

Description of study characteristics

The mean age of the study population was reported in 20 of the studies and ranged between 32 and 51 years, with the minimum age of those treated being 19 years and the maximum age being 72 years (Table 11). 15,31,46,47,49–54,56–65 Only one study appeared to put an age limit on study inclusion,48 although tacit age restrictions may have been applied in other populations but not reported. Sixteen studies reported the reproductive status of the treated women in some format, usually as the mean parity, which ranged from 0.9 to 3.5. 31,46,47,49–54,56–58,60,61,63,65 Five studies clearly did not include any nulliparous women,46,51,53,65,66 whereas in the study by Tropeano et al. 52 nearly half of patients were nulliparous. There was no consistent reporting of any other demographic data, such as body mass index, or of history, such as the duration of symptoms.

Bilateral embolisation was conducted on 478 women, while 384 had only left-side embolisation and six only had right-side embolisation. For the remaining 439 women, it was unclear whether unilateral or bilateral embolisation had been performed. The total number of veins embolised was also unclear, as the procedure was frequently reported as being performed on all incompetent veins including adjacent branches. Furthermore, the number or proportion of women who had the OV or IIV embolised is undeterminable. One of the larger series of 218 women47 reported the distribution of the 526 veins treated as 27% right IIV, 23% left IIV, 7% right OV and 32% left OV, but the degree of multiple vein embolisation could not be extracted from the study report.

Embolisation of pelvic veins was achieved using a sclerosant in 229 patients, by the use of metal coils (stainless steel or platinum) in 660 patients and by a combination of both in 405 patients, with the method unclear in the remaining 13 patients. The sclerosant used varied considerably and included 1–3% sodium tetradecyl sulphate,53,56,59,102 2–3% aetoxisclerol46,52,55,61,66 or enbucrilate. 50,63 There was no apparent trend towards or away from using either technique, either over time or between countries.

Quality assessment

There were 20 case series included in the review (Table 12), with one unable to be reliably assessed owing to uncertainty after its translation. 48 Although the aims and embolisation techniques were clearly stated, one-third did not clearly describe the intended outcome measures and how they were to be collected. Although we sought only prospective trials, it was still not clear in 40% that consecutive patients were included; this was reinforced by poor descriptions of the criteria used to refer patients for venography and inadequate reporting of losses to follow-up. All but one61 were single-institution studies, but none reported conflicts of interest or funding from interventional radiology companies.

Poor reporting hindered the assessment of the risk of bias in the sole randomised trial, but this was likely to have been subject to selection, performance, measurement and attrition biases. The trial compared embolisation with hysterectomy with bilateral or unilateral oophorectomy and, therefore, could not be blinded. The randomisation and allocation concealment method were inadequately described and produced unequal groups, from which there were substantial exclusions from the hysterectomy groups. Overall, the trial was deemed of low internal validity.

Symptomatic improvement

Subjective symptom relief was reported on ordinal scales of complete, moderate/partial and no relief of CPP symptoms in six studies,48,50,53,57,58,63 four of which were definitely prospectively recruited. 50,53,58,63 Early reporting of complete symptom relief, < 2 months, ranged from 33%48 to 80%53 of study participants. Cordts et al. 58 found that 89% of women reported a greater than 80% immediate improvement, although the time scale was not defined, while all six participants in the study by Tarazov et al. 57 reported complete resolution of symptoms at 4 weeks.

Pooling rates of complete, excellent or moderate improvement from these studies gave an overall rate of 75% (95% CI 64% to 85%, I² = 42%) at 4–8 weeks post procedure (Figure 3).

FIGURE 3.

Meta-analysis of rates of complete, excellent or moderate improvement 4–8 weeks following embolisation.

Few studies reported symptom at two time points. van der Vleuten et al. 50 assessed all women twice and found that 67% of women had moderate or obvious improvement at 2 months postoperatively, a figure that increased to 76% at an average of 18 months (±12 months) later. Other studies reported that 85% of women had some symptom improvement at 6 months,52,55 which in the smaller study increased to 95% after 12 months. At an average of 45 months’ follow-up, relief rates of over 80% were reported by two studies15,60 but in a third study sustained relief was reported in only 47% of patients. 49

Studies reported treatment failures as residual symptoms, as unsatisfactory improvement or as the number of repeat embolisations performed. In a larger study, 22 out of 193 (11%) of women had mild or moderate residual symptoms at 6 weeks,47 while another reported that only 2.2% found their improvement unsatisfactory. 48

D’Archambeau et al. 62 reported reintervention by 3–6 months owing to unresolved symptoms in 5.7% of patients, a figure lower than the 15% at 3 months cited by Tropeano et al. 52 There were two repeat embolisations at 6 and 22 months noted in a study of just nine patients,58 while Capasso et al. 63 reported that 5 of 19 patients (26%) underwent repeat embolisations at an average of 15 months.

No study reported quality of life using a generic instrument. Chung and Huh deployed a social readjustment rating scale103 to compare stress between women who had embolisation and those who had hysterectomy, but found no significant differences. 31

Pain scores

Nine studies report on pelvic pain scores using 0- to 10-cm visual analogue scales, although at varying time points after embolisation. 15,31,38,50–52,59,64 The mean pain score was reported in all but one study,52 which reported the median score, and five studies reported a SD around the mean. In all cases, the pain score reduced significantly from a baseline of about 5 to 8 points to between 1 and 4.2 points at 3 months, or between 1.2 and 3 points at 12 months.

Five studies undertook paired t-tests of pain scores before and after embolisation and reported the reductions as being statistically significant, with sufficient data to allow verification by the reviewers. 15,49,51,54,56 Not enough studies had all of the necessary data to allow a meta-analysis to be performed (Table 13).

The decreases were generally sustained in those trials that reported multiple time points (Figure 4).

FIGURE 4.

Time course of pain scores following embolisation for PVI.

Several studies reported on different pain symptoms. Kim et al. 15 noted statistically significant score reductions in pain on standing, lying down, dyspareunia and dysmenorrhoea, as well as a reduction in urinary frequency and the amount of pain relief required, while Venbrux et al. 64 also observed similar reductions in all symptoms except dysmenorrhoea. Gandini et al. 56 reported statistically significant reductions in dyspareunia, dysmenorrhoea and urinary frequency; Creton et al. 65 also considered pain at the site of pelvic varices occurring specifically before or during menstruation, which decreased from a mean of 6.0 to 1.7 at 45 days post embolisation and remained low. van der Vleuten et al. 50 reported on the widest range of symptoms, observing statistically significant improvements in scores at 2 months and an average of 18 months (SD 12 months) for dysmenorrhoea, dyspareunia, worsening of symptoms with walking, standing or sitting, varicose veins and pain in varicose veins. Haemorrhoids improved significantly over long-term follow-up; however, no improvements were seen for urinary symptoms or backache. A composite score of all 10 symptoms, each ranked for severity at baseline and improvement at follow-up, from 1 (no complaints) to 5 (many complaints/worsening), was derived. The mean score improved from 26 before embolisation to 21 at 2 months and 19 at the long-term time point, both statistically significant decreases from baseline.

Impact on the pelvic vein reflux and diameter

Only two studies quantified the diameter of the pelvic veins before and after the procedure. 59,61 One study measured the mean diameter of varices in supine patients, from 6.3 mm to 4.4 mm and from 4.5 mm to 3.1 mm for the left and right OVs, respectively, both statistically significant reductions. 59 In another study, left and right OV diameters reduced from a mean of 6.9 mm (SD 2.1 mm) and 5.1 mm (SD 1.4 mm) to < 4.5 mm on both sides in all cases. 62 Pieri et al. 59 noted that symptoms persisted in women in whom OV varicosities, although decreased in diameter from pre-procedure measurements, remained > 5 mm at rest.

Ratnam et al. 47 was able to perform a repeat TVUS at 6–8 weeks in 193 patients, observing residual mild reflux in 16 (8.3%). Nine women had a second procedure dictated by the ultrasound: six patients owing to moderate persistent reflux and three because of new reflux. Tropeano et al. 52 also repeated the TVUS at 3, 6 and 12 months and observed an absence of reflux and reduced (< 5 mm) pelvic vein diameter in 17 out of 20 (85%) women. The remaining three patients who showed recurrence or persistent left-sided reflux, and were also those who did not report a symptomatic improvement, had a successful repeat procedure performed at 4–5 months and were reported as symptom free after a median follow-up of 15 months.

Impact on varicose veins of the leg

Meneses et al. 51 reported on the impact of pelvic vein embolisation on both pelvic and leg symptoms in women with recurrent varicose veins and clinical and radiological evidence of PVI, using the venous clinical severity score (VCCS). 104 This scale affords four levels (summarised as absent, mild, moderate and severe) to nine criteria (pain, number of varicose veins, venous oedema, pigmentation, inflammation, induration, number, duration and size of active ulcerative lesions) and one therapy criteria (none to continually) to give a 0–30 scale score. The study reported a statistically significant decrease in the VCCS score from 8.4 to 4.0 (p < 0.001) after embolisation in its 10 patients.

Impact on future fertility, menstruation and ovarian reserve

Venbrux et al. 64 captured information on menstruation between 6 and 24 hours post procedure, in 24 of 56 participants, finding no significant difference in the interval or length of menstruation compared with before the embolisation. Kim et al. 15 measured follicular-stimulating hormone, estradiol and luteinising hormone at baseline, 6 months and 12 months and reported no statistically significant differences before and after the procedure.

No study explicitly included or excluded women who desired a future pregnancy or specifically mentioned active follow-up of future pregnancies, and so reports of pregnancies are not likely to have been systematically collected. Three studies15,63,64 reported six successful pregnancies.

Adverse events of embolisation

Six studies did not report any adverse events in their population, although it is not clear whether this equates to technical success in all cases. 46,51,52,56,57,61 Of the remaining 938 women, in total there were 10 cases of vein perforation causing extravascular leakage of contrast media during the insertion of coils. Transient pain following the embolisation was reported in between 8% and 100% of cases and appeared to occur only in the studies using sclerotherapy, either as the sole embolisation method or in conjunction with coils. One large study of 239 women described early adverse symptoms as a ‘post-embolisation syndrome’, reporting that 129 (54%) had transient gluteal or lumbar pain, 61 (26%) had general achiness, 30 (12%) had transient fever (≥ 38 °C) and 21 (9%) had mechanical superficial phlebitis at the point of venous access in the arm. 55

Coil migration after placement was reported in 11 cases; displacement was to the lung in eight cases and the renal vein in two cases, and there was one report of coil protrusion into the femoral vein. In all cases, the coil was retrieved using a catheterised snare, without any lasting harm to the patient.

Individual patient data meta-analysis to explore predictors of effectiveness

One of our stated objectives in this project was to collect IPD from available studies involving embolisation/sclerotherapy in order to identify factors associated with successful outcome, and to perform IPD meta-analysis if possible.

The collection of patient-level data has numerous advantages, including the ability to conduct improved data verification and perform a more comprehensive analysis, including subgroup analysis. 105 However, given the issues outlined earlier in this chapter, we made the decision not to undertake the completion of this objective. These issues included the general low quality of the studies (only one RCT with description of other case series indicating a less than comprehensive design); high variation in interventions and techniques; high variation in the use of outcomes and assessment times; and great inconsistency in the reported characteristics of study participants. IPD meta-analysis is highly resource intensive and we needed to consider the potential cost–benefit of any such undertaking. Our conclusion was that it would be almost impossible to be able to collect and merge enough comparable, high-quality data to make any sophisticated analysis worthwhile.

Summary of review findings

This systematic review of embolisation for PVI found no high-quality studies, and so all estimates of effectiveness are derived from presumed prospective case series reporting on 1308 women, the majority of whom were of reproductive age and parous. Approximately one-third of cases clearly had bilateral embolisation, with metal coil placement being the dominant technique.

Early substantial relief from pain symptoms was observed in approximately 75% of 162 patients in six case series, which generally increased over time and was sustained. Where pain was measured on a visual analogue scale, statistically significant reductions following treatment were observed in all studies. Reintervention rates were generally low. Where measured, embolisation reduced the diameter of dilated veins to a significant degree, with minimal residual reflux.

There were few data on menstruation, ovarian reserve or fertility, but no concerns were noted. Transient pain was a common occurrence following foam embolisation, while there was a < 2% risk of coil migration.

Strengths and limitations of the review

The review followed a registered protocol and focused on a clear question for the assessment of effectiveness. A comprehensive search strategy was constructed and used to screen the widest range of bibliographic sources, with selection undertaken by two reviewers. All outcomes identified a priori were reported on, to the extent that data were available. The embolic techniques employed were generally well described, and follow-up data were available for the majority of participants.

The most significant limitation, which prevents firm conclusions being drawn from the data and recommendations being made, is the quality of the studies identified. There is only one RCT of embolisation in the literature, with hysterectomy as a comparison, which was in itself not free of potential biases. Other randomised trials may have been attempted and not published, introducing publication bias, but unless these were of higher quality than that of the study of Chung and Huh,31 they would have had little impact on our conclusions. The majority of the studies were relatively small case series, with the inherent high risk of bias, no comparative group and frequently ill-defined inclusion criteria. We attempted to restrict the studies included to those where participants were prospectively enrolled, to reduce selection bias, but in some studies it was impossible to be completely certain from the methodology that included participants were not retrospectively identified from medical records.

The effect of embolisation on pain was generally described either in terms of symptomatic improvement or as pain scores, reducing the amount of information available for either outcome. We performed a meta-analysis to get a pooled estimate of improvement rates, but may be criticised for doing so using such low-quality data. Few studies sought pain data on individual pain symptoms, such as dyspareunia or pain on standing, preventing reflection on whether or not embolisation reduces specific symptoms thought to be particularly indicative of pelvic vein congestion. The majority of studies gave a description of adverse events, but the long-term impact in fertility was barely considered.

Given the general low quality of studies and extreme variability of patient-reported outcomes, assessment points and participating characteristics in this systematic review, we declined to attempt to collect IPD from the included studies in an attempt to improve the analysis. We may be criticised for this omission, as it is possible that the review would have been enhanced from capture of patient-level data, but we believed that the cost–benefit of such an undertaking would have been too great for this to be worthwhile.

Interpretation

The pooled estimate of moderate to complete symptomatic improvement following embolisation is consistent with a previous review, as are the rates of coil migration, although this too draws on low-quality case series. 28 Technical success is high, although dependent on pelvic anatomy, and this technique has been widely adopted around the world, as evidenced by the distribution of studies. Few studies addressed the secondary outcomes of the impact of embolisation on menstruation and fertility; these aspects seem not to be discussed in the literature, which is surprising considering that the target population for pelvic vein embolisation is women of reproductive age. This contrasts starkly with uterine artery embolisation for the uterine fibroids, about which there is considerable debate regarding the impact on ovarian function and pregnancy rates. 28,106,107 This may simply be because uterine artery embolisation has the potential to disrupt blood flow to, rather than away from, the reproductive organs, but further data are required to reassure women considering the procedure.

Although the data appear supportive, the quality of evidence is poor. Under the GRADE (Grading of Recommendations Assessment, Development and Evaluation) criteria, the methodological quality is very low and there is no direct comparative evidence against no embolisation. 108 Some reflection on the only RCT is warranted. The study population was thoroughly evaluated to exclude those whose pain could be attributable to organic pathology or who responded to medroxyprogesterone treatment, and who did not wish to retain their uterus and ovaries. The 118 patients were randomised to either embolisation or hysterectomy with bilateral oophorectomy, plus hormone replacement therapy, or hysterectomy with unilateral oophorectomy, on the predominantly congested side. Patients rated their pain on visual analogue scales at 3-monthly intervals after their procedure, and were also stratified according to their stress scores over a recall period of 12 months prior to randomisation. 103 The study reported that the embolisation group and hysterectomy with bilateral oophorectomy group had statistically significant improvement in pain scores, whereas those with only unilateral oophorectomy did not, but these assertions could not be replicated. Furthermore, the method of randomisation was unclear and there were substantial post-randomisation exclusions, which undermine the credibility of the trial results.

We were unable to investigate the heterogeneity of results across the studies other than for the outcome of moderate to complete symptomatic improvement, reported in six studies, where moderate heterogeneity was identified. 48,50,53,57,58,63 This could arise from arise from a variety of sources, such as from the embolisation technique, the way in which the question of improvement was presented and categorised or from the population included in the study. We included studies only that intended to report on pain or symptom improvement and yet were able to directly compare outcomes in either the short or medium term in only a minority of studies, which is a concern. We cannot be sure that some studies did collect data on these outcomes, but did not report the outcomes if they failed to show statistical significance or consistent results.

The precision of the estimates for this outcome is moderate owing to the relatively small number of studies and participants per study, with a CI of ± 10% around the pooled rate of substantial symptomatic improvement of 75% in the short term. In one of the largest series, 50% of participants had total relief and a further 37% had partial relief at 6 months,55 which was consistent with other studies reporting in the 6- to 12-month time frame. 54,62 This would probably encompass a range of response rates that would be acceptable to women, given the low incidence of adverse events. However, we cannot be certain that publication bias does not exist and that this result is not an overestimation of the true effect, with less incentive for unremarkable results to be reported and commercial pressures not to report poor outcomes. Randomised trials are now almost routinely pre-registered in open-access databases, allowing an assessment of the degree to which there are unreported data, but no such mechanism exists for case series.

If we can accept that embolisation provides pain relief, despite the limitations of the literature, the next questions are whether or not particular presenting characteristics can predict a successful outcome, and what the optimal technique is. A retrospective case series of 41 women access the relationship between technique (bilateral or unilateral), parity and location of varices (thigh or labial compared with pelvic) and clinical outcome. There was no statistically significant predictor of a successful outcome among these variables, although a trend towards a higher rate of success in grand multiparous women was noted. No association between these variables and treatment failure was observed, but in both circumstances it may just be that there are too few patients for a small association to be observed. 98

Occlusion of pelvic veins is achieved by means of metal coils, a sclerosant or a combination of the two, according to the radiologist’s preference, with no apparent trend for any technique emerging. Stratification of the meta-analysis of symptomatic improvement by coil or sclerosant is limited by the number of data, but seems to suggest that the overall rate of substantial improvement is reflective of the studies using sclerosant at about 75%, whereas the two studies using solely coils had higher rates, at 89% and 100%, respectively. 57,58 No obvious differences between methods on the reduction in pain scores were seen. In the absence of randomised comparisons, there are no data to indicate superiority of any embolic agent. It may indeed be that the presentation of the varices and anatomy of the veins are the key determinants of successful elimination of reflux, with perhaps both methods being required for the most dilated veins.

Complications of embolisation appear to be limited to short-term pain and fever in a reasonable proportion of sclerosant cases, or an uncommon incidence of coil migration. Coil placement is a relatively straightforward procedure but may be subject to recanalisation or development of collaterals, as has been observed in male varicocele. 109 Some radiologists prefer liquid sclerosant, which can reflux into any collateral veins, owing to its localised effect, the perception that a more extensive embolus is produced, and also the cost compared with metal coils. 98

Further questions remain regarding the embolisation technique. Should only veins observed to be refluxing be targeted or should bilateral embolisation be the default approach? Should only the OV be embolised, or should the IIV be treated too, and, if so, should this be in the same procedure. The data in our selected studies do not help us to address these controversies. The early experience was mainly of selective embolisation of the left OV,57,63 with bilateral procedures becoming more frequent over time. It is not possible to consistently determine from study reports just how many of the later cases of unilateral embolisation were intended or were as a result of difficult access to the right OV. One larger series employed a deliberate strategy of initial OV embolisation followed by repeat venography of the IIV and embolisation where indicated, with 85% having this second embolisation. 15 Pain relief and symptomatic improvement, though, are subjective experiences, and will not be perceived and reported according to the location of the reflux, meaning that it is unlikely any relationship between laterality and improvement would be seen.

Populations

The first survey took place in May 2013, during the 1st World Congress of Abdominal and Pelvic Pain conference in Amsterdam, the Netherlands. This was the inaugural congress of the Abdominal and Pelvic Pain Specialist Interest Group of the IASP and was attended by over 500 abdominal and pelvic pain professionals from 10 disciplines and no fewer than 46 countries. The congress organisers could not provide the e-mails of the attendees, as consent had not been given for contact by third parties, or provide conference bags to attendees. Therefore, to promote the survey, A5-sized invitation postcards were handed to people as they were exiting a plenary session and were also strategically placed in the congress centre. The card gave a brief statement about the purpose of the survey, a weblink to the online survey and a QR code (quick response code) to facilitate completion of the survey on a smartphone. It is estimated that approximately 300 postcards were handed out or taken.

The second survey was sent to consultant members of the British Society of Interventional Radiology in October 2014 by e-mail, from the society secretariat, as again their membership had not given consent for third party contact. This contained a weblink to an online questionnaire and stated that the survey was part of a National Institute for Health Research research project. The e-mail was sent to approximately 600 e-mail accounts, although how many were received and opened is unknown.

The third survey was sent to female members of the Pelvic Pain Support Network (PPSN), who were not health professionals and had, at some point, asked to join the mailing list of this group. The weblink to the survey was distributed in October 2014 from the group’s e-mail address to approximately 800 members. The invitation and survey title did not include the phrase ‘pelvic congestion syndrome’, but did state that the intention was to explore understanding of a rare cause of CPP.

Questionnaire framework

Each questionnaire was tailored to the aim and the target population, with 15, 11 and 7 questions in total in the three surveys, respectively. The surveys were undertaken through Surveymonkey^TM (www.surveymonkey.com; Palo Alto, CA, USA), an online survey development website that enables users to create and publish professional questionnaires online. The package used enables in-built logic, whereby the answers to a question can direct the respondent to particular subsequent questions. It allows respondents to reverse through the questionnaire to edit responses, while making some answers mandatory.

Questionnaire items

The individual questions for the three surveys are shown in Tables 14–16, respectively. Questions were closed, with single answers, multiple answers or ordinal frequency responses (usually, sometimes and never). Two radiologists piloted the second survey, while the chairperson of the PPSN reviewed the third. Not all questions in the survey were mandatory; those that are have a footnote marker in Tables 14–16. Free-text fields were provided for responses not falling within our categories and for general comments.

In the survey to the support group members, the final question, regarding the need for further research, was contextualised with the following preamble:

Embolisation is the first choice of treatment for PCS. There is a wealth of data from women who have had this procedure, of which 80% have reported an improvement in pain following embolisation. Even though there this overwhelming evidence, NICE [National Institute for Health and Care Excellence] and other professional bodies representing gynaecologists and radiologists do not recommend the use of pelvic vein embolisation. This is because they do not feel the evidence base on embolisation, compared to other or no treatment, is strong enough.

Data synthesis

SurveyMonkey provides analytical tools in the package which allow the option to present data as histograms or export to a Microsoft Excel^® (Microsoft Corporation, Redmond, WA, USA) spreadsheet for further manipulation. Results are presented here as percentages and frequencies, with histograms for some questions.

Survey 1

A total of 18 responses were received, giving an estimated response rate of 6%. The majority of respondents were gynaecologists or gynaecological surgeons (44%, 8/18), followed by pain specialists (28%). Other specialists completing the survey included urologists/urology specialists (17%) and sexual health specialists (11%). Respondents were mainly from the UK or the rest of Europe (83%, 15/18); the remainder were individuals from Australia and Philippines. One respondent failed to disclose their location.

Diagnosis of pelvic congestion syndrome

When those surveyed were asked what the most important observations were for a diagnosis of PCS, the following four symptoms were the most popular answers (respondents were asked to select as many as they believed applied): dull pain exacerbated by standing up (33%, 6/18), dyspareunia (22%, 4/18), dilated OVs (28%, 5/18) and dilated pelvic veins (39%, 7/18). Other selected observations included tortuous/twisted venous plexuses (11%, 2/18), refluxing and incompetent OVs (6%, 1/18) and refluxing and incompetent pelvic veins (17%, 3/18). Irrelevant observations, not selected by any respondent, were visible vulval or groin varices, ovarian and renal vein compression by superior mesenteric artery and iliac vein compression against the spine or pelvis. One respondent reported, as free text, a patient having difficulties during sex due to too much lubrication.

There were two dissident respondents. One stated that had not diagnosed a single case of PCS in 10 years in the pelvic pain field and had only very rarely seen dilated veins at laparoscopy, and they believed these to be coincidental. Another was not convinced that PCS, as a condition, even existed, believing that symptoms were due to endometriosis or adenomyosis not visible at laparoscopy.

In terms of the imaging methods used to diagnose PCS, respondents were asked to rate their usage of each test, as illustrated in Figure 5. Eight respondents skipped the question completely (with one describing it as not applicable) and, as responses were not mandatory, the number of responses per test varied from three to eight.

FIGURE 5.

Survey 1, question 4: which imaging methods do you use to diagnose PCS?

Transvaginal ultrasound with Doppler and laparoscopy were the most frequently used techniques. Four people usually used laparoscopy and two used it sometimes; this was the same for TVUS with Doppler, with four using TVUS alone (three usually and one occasionally). One or two individuals did not use these three techniques. Three people used TAUS.

One respondent said that he or she frequently used MRI, and another used this occasionally, while only one respondent used MR venography, and then only occasionally. Only one respondent used CT, and this was frequent. Fluoroscopic percutaneous or transfundal venograms were not used by any respondents.

Among those who specifically reported avoided using an imaging method, four never used TAUS, MR venograms and CT, while one did not employ laparoscopy.

Survey 2

Survey 2 was initiated by 63 respondents, giving an estimated response rate of 10%. The initial screening question asked whether or not radiologists were currently performing pelvic vein embolisation in women with CPP and PVI; 47 out of 63 (75%) respondents answered positively. The remaining 18 were diverted to the final question.

Twelve respondents made additional comments to accompany this question. Three did not currently undertake embolism, although one had done this previously and another indicated that it was done within his or her department; the third stated that he or she ‘could not get my clinician colleagues to believe there is such a thing’ as PCS. Of those respondents who did currently undertake embolisation, at least five reinforced the infrequency of this procedure, although one stated that the number of women he or she saw had been slowly increasing over time. Three radiologists related their embolisation practice to the treatment of varicose veins. One used embolisation for varicose veins in the leg, which originated in the pelvis, and noted that although most of these patients had PCS, they were not referred specifically for that reason. Another commenter echoed this sentiment, saying that for them, pelvic vein embolisation was usually for leg varicose veins and/or vulval varices, and rarely for PCS alone. A third respondent stated that his or her largest group of patients had PVI associated with lower-limb varicosities. A further respondent highlighted the need for a multidisciplinary team approach, which would include a vascular surgeon (with an interest in PCS), a vascular ultrasound specialist and a gynaecologist.

Diagnosis of pelvic congestion syndrome

The 45 respondents who undertook embolisations were asked to subjectively choose all of the symptoms and observations that they felt were important in order for pelvic vein embolisation to be considered (Figure 6).

FIGURE 6.

Survey 2, question 2: to you, what are the important observations for consideration of pelvic vein embolisation?

With regard to patient-reported symptoms, 32 out of 45 respondents (71%) regarded dull pain exacerbated by standing up as an important observation to necessitate embolisation, while 27 respondents (60%) considered dyspareunia important. Dilatation of the OV was a criterion for 32 (71%), but only one-quarter thought the same for dilated IIVs. Only 16 (36%) considered tortuous pelvic veins important. With regard to refluxing and incompetent veins, around two-thirds rated the OV and other pelvic veins as crucial, but this was surpassed by the 37 (82%) who considered visible vulval or groin varices important. The radiological condition of Nutcracker syndrome was generally discounted. The median number of signs and symptoms considered as indications was five.

Other important indications suggested by respondents in open responses included irritable bowel syndrome, bladder instability, labial varicosities, negative laparoscopy findings for other causes of pain and non-cyclical pelvic pain. The relationship with varicose veins and communicating pelvic veins was reiterated. Another respondent related the criteria they used to those they applied to treating a varicocele in men, during which treatment they embolised any gonadal vein that was distended and could be creating venous pressure into the pelvis.

In terms of the imaging methods used to diagnose PCS, respondents were asked to rate their usage of each test, as illustrated in Figure 7. As a response for each test was not mandatory, the number of responses per test varied from 8 to 36 of 45 respondents.

FIGURE 7.

Survey 2, question 3: which imaging methods do you use to identify pelvic vein reflux before discussing embolisation with the patient?

Magnetic resonance venography was the most frequently used technique, with 17 of 36 (47%) respondents usually employing it, 16 using it often and only 3 never using it. Traditional venography was the most consistently used technique, with 22 out of 34 respondents stating that they used this method. Ultrasound with Doppler was preferred to simple ultrasound, with transvaginal and transabdominal scans being equally popular and comparable with CT. Transfundal venogram appeared to be an obsolete technique.

The responses to the open-ended questions suggested there could be a staged approach to diagnosis, reflecting the clinical situation in which venography immediately precedes embolisation. One respondent indicated that he or she would choose TVUS with Doppler as their first-line imaging test; another aimed to perform Doppler ultrasound of the leg veins, which, if it showed that they originated from the pelvis, would prompt a fluoroscopic venogram, with consent from the patient to perform embolisation if appropriate. Another respondent held conflicting views, stating that he or she found non-invasive imaging methods unreliable and would instead proceed straight to conventional venography.

Number of procedures performed

Among the 45 radiologists who undertook pelvic vein embolisations, the number of procedures performed over their careers ranged from 1 to 1100. The majority had performed fewer than 20, but 4% had performed over 500 (Figure 8).

FIGURE 8.

Survey 2: distribution of number of procedures performed by interventional radiologists in their career.

Effectiveness of pelvic vein embolisation

This question was asked to estimate the range of radiologists’ beliefs about the effectiveness of embolisation. Respondents were asked, if embolisation was performed using ideal technique at the hands of an experienced radiologist, in patients with clearly defined PVI, how they would rank the probability of the following four outcomes, collected from the patient at 6 months, and then compared with pain before the procedure. The four outcomes were complete or significant reduction in pelvic pain, moderate reduction in pelvic pain, very small reduction in pelvic pain and no reduction in or worsening of pelvic pain. Respondents were asked to assign a percentage probability to each outcome, ensuring that the total added up to 100%. Responses are presented as weighted average percentages in Figure 9.

FIGURE 9.

Survey 2: distribution of beliefs regarding the effectiveness of embolisation for PCS.

Survey 3

From the e-mailshot, 69 women commenced the survey, representing a response rate of approximately 8%. We first ascertained how many respondents were familiar with the term ‘pelvic congestion syndrome’ as a cause of CPP in women. Our results found that fewer than half of the women had knowledge of PCS (32/69, 46%). As the remaining questions were designed to explore women’s understanding of PCS, those who were unaware of PCS were diverted to the end of the survey, reassured that PCS was an under-recognised and rare cause of CPP and thanked for their contribution. Of those who were aware of the condition, 3 out of 32 (9%) had received a diagnosis of PCS as the cause of their pelvic pain, with one woman having been told that PCS may be the cause of her pain. Other than one respondent who did not have pain, the remainder either had or at one time had pelvic pain, but they had not been diagnosed with PCS.

Symptoms and patient characteristics of pelvic congestion syndrome

Figure 10 shows the distribution of responses from 30 women, when they were asked which symptoms they thought were indicative of PCS. Six respondents admitted that they did not know, while the remainder selected an average of five symptoms. Symptoms were described in lay terms, which we also use here.

FIGURE 10.

Survey 3: symptoms considered indicative of PCS by women with CPP.

The symptom identified by the most respondents (21/30; 70%) was a dull pain made worse by standing for a long time. The next most frequent symptoms, each reported by 43–46% of women, were pain between periods, pain during periods and pain inside during penetrative sex. The remaining symptoms were selected by between 20% and 27% of respondents.

To assess their perception of the risk factors, respondents were asked for the characteristics of women who they felt were more likely to have a diagnosis of PCS. Over half of the 30 who responded were unsure. All of the remaining 13 believed that PCS was a condition of women of reproductive capability. One thought that nulliparity was a risk factor and two identified grand multiparity, and while three thought that vaginal delivery was important, none thought that PCS was associated with caesarean delivery. Three also thought being overweight was a risk, but that ethnicity was not. Four qualified their choices by stating that PCS was a condition that any woman could develop.

Key findings

We undertook three surveys to get a broad perspective of current understanding and management of PCS, eliciting responses from those with a specialist interest in pelvic pain, from interventional radiologists and from women with CPP. Although this work predominantly focuses on the UK, we believe that it represents the only recent and relevant review of management of PCS, beyond the case series discussed in Chapter 6.

A few key themes can be drawn out from the surveys:

• There is variation between pain specialists and interventional radiologists in their approach to diagnosis.

• Surgical management does not appear to be favoured.

• The majority of interventional radiologists do not perform many, if any, embolisations for PVI, and although their confidence in the procedure is reasonably high, there is a desire to strengthen the evidence base.

• Even among women with CPP, fewer than half had any knowledge about PCS.

• There remain some dissenting opinions regarding PCS as a cause of CPP.

Strengths and limitations of the surveys

Surveys can be unrepresentative if they collect information from only a limited number of people and if those who do respond are not representative of their constituency. Unfortunately, all three surveys were limited in size, both initially in the number of people connecting to the SurveyMonkey questionnaire and in those who answered affirmatively to the first question of the second and third surveys, which was designed to select those with experience or understanding of PCS. As discussed before, it was not possible to compare any characteristics of responders and non-responders among the British Society of Interventional Radiologists membership, so it is likely that the respondents are those for whom embolisation is more pertinent. The low response rates to all three surveys are of concern, but not unique. A recent web-based survey of the membership of the European Society of Gynaecological Endoscopy, regarding laparoscopic morcellation practice, generated a response rate of around 5% (ESGE Secretariat, 2014, personal communication). There was no monetary incentive for completing the survey; vouchers for credit at online retailers are increasingly used to boost response rates. The second and third surveys were open for response for only 2–3 weeks, with no reminder sent to non-responders, although the SurveyMonkey analytics show that the peak time for responses was within a few days of the e-mailshot, so further reminders would have been likely to give diminishing returns.

Short surveys are more likely to be completed in full. We could see some diminution in the number of respondents in later questions in each survey, possible reflecting question fatigue. For the questions in which frequencies were ranked, in retrospect it would have been beneficial to require a response for each queried test or treatment, so that the relative proportions choosing usually, sometimes or never always totalled the same. We can only infer that those not providing a rating for each test or treatment had never used that test or treatment. The ability to skip past questions was useful in the first survey. We did not know the spectrum of professions who would respond when we designed the survey, so we attempted to cover a broad range of management aspects. The terminology used in the later surveys was improved as a result of feedback from the first survey.

Online surveys are undoubtedly a popular modern method for collecting data and provide advantages such as instant analysis of the data. They are, however, restrictive in the type of questions that can be asked, to some extent. We were not able to elicit respondents’ beliefs about the effectiveness of embolisation in a way that would provide a subjective probability for a range of potential effects. Previous methods of numerically representing beliefs have involved respondents being provided with a range of outcomes and a visual analogue scale for each outcome. 110 In order to apply this technique in our situation, we would have had to present a list of perhaps 11 options on a spectrum ranging from, at one end, embolisation providing 100% symptomatic improvement, to 80% improvement, to 60% improvement, to no benefit over no treatment, and finally to the other extreme of possible treatment effects, namely embolisation worsening the pain. Respondents would use the visual analogue scale to mark their subjective estimation of the probability of each outcome, on an anchored line from impossible, to increasingly likely, to certain. Previous users of this approach have tended to undertake this exercise while the respondents are present, for example at a meeting; this allows the users to explain the objective of the survey and provide instructions to the respondents on how to complete the visual analogue scale.

However, although marking such a scale on paper is easy, SurveyMonkey does not provide the equivalent of a visual analogue scale. The approach we chose was an adaption of a weighting method,111 albeit with four non-numerical categories. We asked radiologists to assign a distribution of probabilities, forcing the responses to sum to 100%. Our primary outcome was symptomatic improvement, which is a subjective outcome itself and so would be variably interpreted by radiologist. Despite this, we can report that about 38% of radiologists believed that embolisation would provide a complete resolution of symptoms or significant improvement, whereas 18% believed embolisation would not provide an improvement in symptoms or would worsen them.

Interpretation

Pain specialists and radiologists both considered dilated pelvic veins as highly indicative symptoms of PCS, although even more radiologists appeared to rank visible vulval varices as a relevant observation and one-third of women were aware of this as a symptom. All three groups considered dull pain from prolonged standing to be important, although this is rarely described in the definitions of PCS reviewed in Chapter 3.

Pain specialists and radiologists diverged to a certain extent in their chosen diagnostic methods, which probably reflects their respective positions in the management pathway of women with CPP. The former were more likely to use ultrasound and laparoscopy, procedures that they were likely to be capable of performing and which would also be part of the typical gynaecological investigation of CPP. MR or fluoroscopic venograms were the preferred methods of radiologists, which is unsurprising given that they constitute the definitive method of identification and localisation of PVI and would be undertaken prior to embolisation. Female members of the support group, perhaps influenced by their personal experiences, believed that three or four tests would be required. Although laparoscopy was frequently cited, it is thought by some to underestimate the true prevalence of PVI. It does, however, have its place in eliminating other pathological causes of CPP.

In the limited sample of pain specialists, very few were inclined to offer medical treatments, and there was no clearly preferred treatment. This reflects the fact that women may have already tried various medical treatments to no avail. It has been noted that it can take years for a woman to obtain a diagnosis of endometriosis,7 and it is likely that the same issues apply for PCS, given the uncertainty around its definition. No respondents, including the gynaecologists, proposed surgical treatment. This is not entirely unexpected, as hysterectomy is a major operation and perhaps should not be considered while there are non-invasive alternatives available, although ligation of the ovarian arteries may achieve the same effect as embolisation. Hysterectomy with oophorectomy has been used historically and was the comparator in the only randomised trial of embolisation. 31 This observation also vindicates our decision to exclude pelvic ligation from our review in Chapter 6. Only a minority of pain specialists considered acupuncture and physical and psychological therapies, which was perhaps unsurprising given that for CPP in general there is little evidence for any non-surgical or non-pharmacological interventions. 112

In our sample, the pain specialists did not appear to consider a referral to an interventional radiologist, while most radiologists had performed only relatively few embolisations in their career. From this we cannot infer that a lack of referrals is the factor limiting the number of procedures undertaken, although the uncertainty and controversy around the definition of PCS is considered contributory by radiologists, from their open comments in the survey, and reflects our findings in Chapter 3. In the studies reviewed in Chapter 6, approximately half of the women were embolised using coils, and this method dominates UK practice.

There is a degree of discordance between the extent of symptomatic improvement cited in the observation studies discussed in Chapter 6 and the beliefs of radiologists in the UK. The review revealed a pooled estimate of 75% achieving complete or significant improvement, while radiologists put an average of 38% probability of this outcome. That they anticipated a 20% risk of no benefit from embolisation possibly suggests a belief either that some women are refractory to this treatment or that there is not a complete association between PVI and CPP. It also hints at publication bias in the literature, if these beliefs reflect radiologists’ experiences.

What the distribution of beliefs does clearly show is that there is uncertainty regarding the effectiveness of embolisation; this was supported by the overwhelming majority of radiologists, who thought that there was insufficient evidence and that a randomised trial was needed. This was echoed by the members of the support group, although, given the general lack of awareness of the condition, this may reflect a higher level of knowledge among those who do respond. Challenges facing any potential future research are substantial, given the prevalence, the diagnostic criteria, the attitude of specialists for referral and the potentially limited number of radiologists offering the procedure, but there appears to be an appetite for further research.

Review of diagnostic criteria and definitions

There was no single, clearly defined criterion for a diagnosis that was reported in all of the studies included in the review. The majority of studies cited pelvic pain, dilated OVs and venous reflux or congestion as principal features of PCS, but many did not give thresholds or further clarification or, where they did, were heterogeneous. There is a need for a globally accepted diagnostic standard for PCS, which would help to standardise clinical evaluation and facilitate comparisons of outcomes of treatment effectiveness. We propose a potential template of diagnostic criteria based on the following criteria:

• CPP, considered by the woman to have an impact on her quality of life, > 6 months’ duration, located in the pelvic region below the umbilicus

• delineation of the presence of dysmenorrhoea, dyspareunia and pain after prolonged standing, and a subjective ranking of the severity of each symptom

• presence or absence of visible varices

• presence or absence of pelvic vein variants

• OV dilatation

• reflux in pelvic veins, in terms of retrograde flow on Valsalva manoeuvre, delayed refilling time and filling of contralateral veins.

Further work is required to determine score weightings and cut-off values, and validation of a combined score in a prospective study.

Association review

Overall, there is a lack of well-designed case controlled studies to delineate the relationship between CPP, dilated pelvic veins and PVI. In the five case–control studies with appropriate data,6,16,23,41,77 the associations were generally fairly similar, with three studies showing statistically significant associations (odds ratios of between 31 and 117). 6,16,77 The two smallest studies failed to reach statistical significance in the odds of association, perhaps because they were too small to detect a difference. The proportion of women found to have PVI who reported CPP ranged considerably, from 39% to 91%. Polycystic ovaries were observed more frequently in the group with CPP and PVI in two studies. The prevalence of PVI ranged considerably, although the majority of women with PVI had CPP. Conversely, in the four studies of asymptomatic women undergoing pelvic vein imaging for other reasons, no more than half had PVI, although again the prevalence ranged widely. Where lower-limb venous insufficiency was seen, between 60% and 77% of women also had pelvic varices.

Review of the accuracy of imaging tests

Transcatheter selective venography performed in a tilted position remains the gold standard for the determination of PVI and is an integral part of pelvic vein embolisation, and, therefore, remains an essential part of the diagnostic process. TVUS with Doppler has a place in the screening of women for pelvic varices, although the data on accuracy are limited. One study provided diagnostic accuracy parameters of 96% sensitivity (95% CI 92% to 99%) and 100% specificity (lower 99% CI 97%), with positive and negative predictive values of 100% and 94%, respectively. 36 Similarly, MR venography appears to have good accuracy, but accuracy data are limited to one study, suggesting 88% sensitivity and 67% specificity for identifying PVI in the OV. 38 Imaging modalities that demonstrate both venous dilatation and reflux are necessary.

Review of the effectiveness of embolisation

This systematic review of embolisation for pelvic congestion system found no high-quality studies, so all estimates of effectiveness are derived from presumed prospective case series; therefore, all findings should be interpreted cautiously in view of the high risk of biased contributory data. These series reported on 1308 women, the majority of whom were of reproductive age and parous, while approximately one-third of patients clearly had bilateral embolisation, with metal coil placement being the dominant technique.

Early substantial relief from pain symptoms was observed in approximately 75% of patients, which generally increased over time and was sustained. Where pain was measured on a visual analogue scale, statistically significant reductions following treatment were observed in all studies. Reintervention rates were generally low. Where measured, embolisation reduced the diameter of dilated veins to a significant degree, with minimal residual reflux.

There were few data on the impact on menstruation, ovarian reserve or fertility, but no concerns were noted. Transient pain was a common occurrence following foam embolisation, while there was a < 2% risk of coil migration.

We considered that the proposed IPD meta-analysis would not yield enough comparable, high-quality data to make any sophisticated analysis worthwhile and so we did not proceed with this.

Surveys of clinical practice and lay awareness of pelvic congestion syndrome

We undertook three surveys to get a broad perspective of current understanding and management of PCS, eliciting responses from those with a specialist interest in pelvic pain, from interventional radiologists and from women with CPP. From these surveys, a few key themes can be drawn out, perhaps most significantly that there are some dissenting opinions regarding PCS as a cause of CPP. First of all, there is variation between pain specialists and interventional radiologists in their approach to diagnosis, but it is obvious that surgical management does not appear to be favoured. The majority of interventional radiologists do not perform many, if any, embolisations for PVI and, although their confidence in the procedure is reasonably high, there is a desire to strengthen the evidence base. Even among women with CPP, fewer than half had any knowledge about PCS.

Literature searching

The key strength of this overview is that two extensive, protocol-driven literature searches were undertaken, one specifically for diagnosis and the other for treatment, ensuring that the maximum number of appropriate studies was identified. We searched multiple databases, including those that focus on journals not indexed by MEDLINE or EMBASE, to locate studies not cited by previous reviews, and selection was undertaken by two reviewers. We could be criticised for not conducting a search specific to the review of the association between PVI and CPP, thereby potentially missing important citations, but we are confident that the breadth of the search strategies enabled the capture of all relevant studies. In screening several thousand citations for the accuracy and effectiveness reviews, we located all of the papers included in the association review. Moreover, in the process of reviewing the shortlisted studies for these other reviews, we scanned the reference list for additional relevant papers and did not locate any additional contemporary papers. We are, therefore, confident that we have located all relevant case–control studies, although we may have missed some studies which were primarily focused on lower-limb insufficiency with secondary considerations of the coexistence of PVI.

For the review of definitions, we took a selected subset of all possible studies, selected on the basis that they were included in the accuracy and effectiveness systematic reviews, for which, again, we could be criticised. In the process of screening citations for the other reviews, we did not locate a single publication in which a consensus diagnostic standard had been derived by valid methods and believe it is likely that the identification of diagnostic themes has reached data saturation.

In each review, the required data were prespecified to avoid selective outcome reporting. In the review of diagnostic definitions, we categorised the criteria to predetermined groups and adopted a strict policy of extracting the definition from the methods section of each report. In the accuracy review, we prespecified the reference standard for consistency. Although we initially proposed to include ovarian ligation in the review of the effectiveness of treatments, dropping this modality from the review will have no implications for the conclusions, as ligation is nowadays rarely performed.

Quality of primary studies

A quality appraisal of the selected studies was conducted for the reviews of association, accuracy and effectiveness, using appropriate, validated quality checklists. This enabled interpretation of the data in the light of the potential biases in the primary studies. The generally poor quality of the studies’ methodology and the heterogeneity of studies, together with inadequate reporting of disaggregated data, result in our inability to perform meta-analysis in the association and accuracy reviews, and meant we were able to combine data for only one outcome in the effectiveness review.

For the association review, the six most pertinent studies were unfortunately of mixed methodological quality. 6,16,23,41,77,78 They drew on clinically disparate populations and defined PVI inconsistently, and in the case–control studies there was no matching on important confounders such as parity.

In the accuracy review, there were no concerns about the reference standard used in the selected studies. However, there were several issues common to most studies that undermine their estimates of accuracy. Only 2 of the 12 studies were prospective23,38 and included consecutive women, while in about half of the studies ultrasound was being used as a triage test to identify those women who would go on to have embolisation; consequently, those who did not show evidence of PVI on ultrasound never had the reference standard venography. Only the two studies attempted to blind the index test and only one of these was rigorous in its methods to prevent ascertainment bias in the analysis of the images. 40

The most significant limitation of the effectiveness review is the absence of randomised data. Only one RCT of embolisation is reported in the literature, and this was considered at risk of potential biases. The majority of the studies were relatively small case series, with no comparative group and frequently ill-defined inclusion criteria. We attempted to restrict the studies included to those in which participants were prospectively enrolled, to reduce selection bias, but in some studies it was impossible to be completely certain from the methodology that included participants were not retrospectively identified from medical records in some studies reviewed.

The effect of embolisation on pain was generally described either in terms of symptomatic improvement or as pain scores, reducing the amount of information available for either outcome. Given the general low quality of studies and the extreme variability of patient-reported outcomes, assessment points and participating characteristics in this systematic review, we did not collect IPD from the included studies in an attempt to improve the analysis.

All three surveys were limited in size and so may be unrepresentative. It was not possible to compare any characteristics of responders and non-responders. For the questions where frequencies were ranked, in retrospect it would have been of benefit to require a response for each question, so that the relative proportions were comparable. We were not able to elicit beliefs about the effectiveness of embolisation in a way that would provide a subjective probability for a range of potential effects.

Implications for health care

There exist a proportion of women for whom no cause for their pain can be found at laparoscopy; this causes anxiety for the patient, and the search for a diagnosis can be protracted, placing a significant demand on health-care resources. The strength of the evidence with respect to diagnosis and management is insufficient for any clinical recommendations to be made, but some good-practice points can be listed. When taking a clinical history, gynaecologists should ask about specific pain symptoms, including whether or not pain is more severe after periods of standing and if it is relieved by lying down. During examination of the patient, the presence of vulval and lower-limb varicose veins should be noted. Transvaginal Doppler ultrasound should be made available for women exhibiting symptoms indicative of PVI. If this technique identifies incompetent pelvic veins, the radiologist can discuss the possibility of PCS as a diagnosis with her gynaecologist, highlighting the uncertainty in the data. If there are interventional radiologists available who perform embolisation of pelvic veins, a referral may be considered. Women should be counselled that the embolisation, although apparently safe, may not provide complete relief of symptoms. There are no robust data on clinical effectiveness and cost-effectiveness, so under current guidelines from the National Institute for Health and Care Excellence it not possible to state whether or not embolisation provides value for money.

Recommendations for research

There is scope for considerable further research, with robust methodology and of adequate size, into the condition known as PCS. The question of the association of PVI and CPP requires a well-powered case–control study, in which women with CPP are matched, using age and/or parity, to two or more pain-free controls. All women would need to provide a standardised account of their pain symptoms, their gynaecological and obstetric history, and be examined for vulval or lower-limb varicose veins, before undergoing a consistent TVUS assessment using the most modern Doppler technology. Interpretation of the ultrasound and Doppler data should be undertaken in duplicate, with readers blind to each other’s assessments, to determine interobserver reliability. Ideally, a small subsample of patients should have the same ultrasound procedure performed at two time points, to assess if timing in relation to the menstrual cycle is important, and to assess intraobserver consistency. This would provide data on the odds of PVI being associated with CPP and on the reproducibility of ultrasound protocol.

This potential association study is predicated on there being a clear definition of PVI, but it will also provide data to derive the diagnostic performance of each individual criterion, in terms of the ability to discriminate PCS from pain-free controls. For example, a receiver operating curve for OV dilatation could be produced, using various thresholds to define dilatation, and the optimum cut-off value could be obtained. Those parameters with a statistically significant difference between the CPP and pain-free groups of women would be incorporated into a logistic regression model. The regression coefficients of the best-fit model could be used to weight scores for each criterion before summation to a combined score.

A further study of women who were prepared to have TVUS with Doppler and conventional venography would be required to provide the threshold for the total score, in addition to reliably determining the accuracy of TVUS with Doppler, again mapping the score on a receiver operating curve to determine the cut-off value that provides the optimum clinical performance with respect to sensitivity and specificity. Finally, the scoring system would be validated in a prospective study of women with CPP to determine its accuracy. This would then provide a practical, valid tool for clinical use that would assist in the identification of women for invasive confirmatory venography and, potentially, treatment of PCS.

Assuming a clear diagnostic standard for PCS can be obtained in this method, there remains the issue of whether or not embolisation is an effective treatment. An adequately powered randomised trial is essential to provide the necessary data, but faces methodological challenges. Pain is a subjective phenomenon and prone to measurement bias unless a placebo or sham intervention is provided. In a previous study of a neuroablative technique for CPP, women who were blinded to whether they had the procedure or a sham equivalent all reported an improvement in pain to a comparable extent. 113 A conventional venogram would be required to verify PVI immediately prior to randomisation between a legitimate or a sham embolisation. How the blinding of the woman can be maintained, with the woman only under sedation, is potentially challenging, but it is ethical given its necessity in reducing the risk of bias and the low risk of complications from venography. There are no precedents found in the use of sham embolisation from the equivalent male condition of spermatic vein varicocele, as in this condition the intention is to improve subfertility, whereas pregnancy rates can be objectively ascertained.

Finally, an economic evaluation of the diagnostic pathway and of embolisation would provide the final evidence of whether or not the proposed management strategy is cost-effective. It would be of general benefit in the area of CPP to collect information on overall and condition-specific quality of life and on the quantity and type of health-care resources consumed by women, as there are few contemporary data in this field.

African Index Medicus

Cumulative Index to Nursing and Allied Health Literature

The Cochrane Library

Database of Abstracts of Reviews of Effects

EMBASE

Date range: 1980 to week 9 2014.

1. chronic pelvic pain.mp.

2. CPP.mp.

3. chronic.mp.

4. exp pelvic pain/

5. pelvic pain.mp.

6. (pelvic adj2 pain).mp.

7. 4 or 5 or 6

8. exp pain/

9. pain.mp.

10. 8 or 9

11. 3 and 7 and 10

12. exp dyspareunia/

13. dyspareunia.mp.

14. 12 or 13

15. exp dysmenorrh$ea/

16. dysmenorrh$ea.mp.

17. 15 or 16

18. 1 or 2 or 11 or 14 or 17

19. exp ultrasonography/

20. ultrasonography.mp.

21. exp ultrasound/

22. ultrasound.mp.

23. exp ultrasonic/

24. ultrasonic.mp.

25. doppler.mp.

26. exp sonography/

27. sonography.mp.

28. 19 or 20 or 21 or 22 or 23 or 24 or 25 or 26 or 27

29. exp magnetic resonance imaging/

30. magnetic resonance imag*.mp.

31. exp mri/

32. mri.mp.

33. 29 or 30 or 31 or 32

34. exp tomography, x-ray computed/

35. tomography, x-ray computed.mp.

36. exp cat scan/

37. catscan.mp.

38. cat scan.mp.

39. exp comput*/ adj2 tomog*/

40. (comput* adj2 tomog*).mp.

41. CT.mp.

42. 34 or 35 or 36 or 37 or 38 or 39 or 40 or 41

43. exp angiography/

44. angiogra*.mp.

45. 43 or 44

46. exp phlebography/

47. phlebography.mp.

48. 46 or 47

49. exp venography/

50. venography.mp.

51. 49 or 50

52. 28 or 33 or 42 or 45 or 48 or 51

53. 18 and 52

Index Medicus for the Eastern Mediterranean Region (IMEMR)

Index Medicus for the South-East Asian Region (IMSEAR)

Latin American and Caribbean Health Sciences Literature (LILACS)

MEDLINE [Ovid MEDLINE(R)]

Date range: 1946 to week 3, February 2014.

Pan American Health Organisation

Population Information Online

Scientific Electronic Library Online

Web of Science

Western Pacific Region Index Medicus

Bioscience Information Service Web of Knowledge

British Nursing Index

Searched for: all (pelvic congestion)

Cumulative Index to Nursing and Allied Health Literature

The Cochrane Library

Database of Abstracts of Reviews of Effects

EMBASE

Index Medicus for the Eastern Mediterranean Region

Index Medicus for the South-East Asian Region

Index of Scientific and Technical Proceedings

Pelvic congestion syndrome AND vein

Latin American and Caribbean Health Sciences Literature

Medion

MEDLINE [Ovid MEDLINE(R)]

Date range: 1946 to week 1, March 2013.

Pan American Health Organisation

Population Information Online

Scientific Electronic Library Online

Web of Science

Western Pacific Region Index Medicus