Endometrial sampling in low-risk patients with abnormal uterine bleeding: a systematic review and meta-synthesis

Background One million women per year seek medical advice for abnormal uterine bleeding (AUB) in the United Kingdom. Many low-risk patients who could be managed exclusively in primary care are referred to hospital based gynaecology services. Performing endometrial sampling (ES) in the community may improve care, reduce the rate of referrals and minimise costs. We aimed to search and synthesise the literature on the effectiveness of ES (Pipelle versus other devices) in managing AUB in low-risk patients. Methods We undertook an electronic literature search in MEDLINE via OvidSP, Scopus, and Web of Science for relevant English-language articles from 1984 to 2016 using a combination of MeSH and keywords. Two reviewers independently pre-selected 317 articles and agreed on 60 articles reporting data from over 7300 patients. Five themes were identified: sample adequacy, test performance, pain and discomfort, cost-effectiveness, and barriers and complications of office ES. Results Pipelle seems to perform as well as dilation and curettage and, as well or better than other ES devices in terms of sampling adequacy and sensitivity. It also seems to be better regarding pain/discomfort and costs. However, Pipelle can disrupt the sonographic appearance of the endometrium and may be limited by cervical stenosis, pelvic organ prolapse and endometrial atrophy. Conclusions The current evidence supports the use of Pipelle in the management of low-risk women presenting in the outpatient setting with symptomatic AUB when combined with clinical assessment and ultrasound scanning. However, the implications of its widespread use in primary care are uncertain and more research is required. Electronic supplementary material The online version of this article (10.1186/s12875-018-0817-3) contains supplementary material, which is available to authorized users.


Background
Abnormal uterine bleeding (AUB), traditionally defined as uterine bleeding that is abnormal in volume, regularity, and/or timing [1] is common and affects 14-25% of women of reproductive age [2][3][4]. In the UK, approximately 1 million women seek medical advice for AUB every year, mostly in general practice [5,6] and even though most cases could potentially be managed exclusively in primary care [7,8], AUB is the fourth most common reason for referral to UK gynaecological services [6,9,10]. AUB has a major impact on quality of life [7], leads to 3.5 million days of work absence [11], and generates significant health care costs. Hospital referrals and hysterectomies are the major components of the £65 million/year treatment costs for AUB [10].
Most cases of AUB are benign and amenable to office-based treatments [12,13]. However, patients often present with a myriad of symptoms, and their assessment requires training and expertise [13,14]. The causes of AUB can be summarised using the PALM-COEIN acronym: polyps, adenomyosis, leiomyoma/fibroids, malignancy (and hyperplasia), coagulopathy, ovulatory disorders, endometrial, iatrogenic, and not otherwise classified [1]. Some patients who present with AUB will have endometrial hyperplasia or cancer which is the commonest gynaecological malignancy in the Western world. Even though the incidence rises after menopause, it can occur at all ages and 7% of cases are under 50 [15,16]. This percentage seems to be rising with increasing prevalence of obesity and diabetes [17,18].
In the UK, women with AUB who are deemed at high risk of endometrial cancer such as those with postmenopausal bleeding (PMB) or family history of gynaecological neoplasms, should be referred to secondary care [19]. For low-risk premenopausal women the guidance is not as clear. Although urgent referral is not required [20], national guidelines recommend that endometrial sampling (ES) should be performed in women over [40][41][42][43][44][45] years to exclude cancer [21,22], but they do not specify whether ES should be performed in primary or secondary care [22].
In the UK, ES for AUB patients has not been traditionally undertaken in primary care. For many years, the standard management was dilation and curettage (D&C) in hospital under general anaesthesia [23][24][25]. However, the need for admission and the risks of perforation and haemorrhage made D&C unpopular [23,25] and various ES devices were developed such as the Novak (a silastic cannula with a bevelled lateral opening [26]), the Tis-u-Trap (a plastic curette with suction [27]), the Vabra Aspirator (a stainless steel cannula connected to a vacuum pump [28]), the Endorette (a plastic cannula with multiple openings [29]), the Tao Brush (a sheath brush device [30]), the Cytospat (a polypropylene cannula with a rhomboid head [31]), the Accurette (a quadrilateral-shaped curette with four cutting edges [32]) and the Pipelle, the most widely used device in the UK (a flexible plastic tube with a distal circular port [27]).
We conducted a systematic review of the literature to identify existing evidence about the effectiveness of Pipelle compared with other ES techniques for assessing low-risk women with AUB which could inform the development of new care pathways in primary care.

Why this study was necessary
Endometrial sampling is thought to be a safe and effective method for histological assessment of the endometrium. It is used as an alternative to the more invasive method of D&C. This is the first review to focus on AUB in low-risk pre-and perimenopausal women. We conclude that ES is a valuable tool in the assessment of these patients and that Pipelle is the best outpatient device available. The evidence supports the use of Pipelle in the outpatient setting but more research is required to assess its impact if introduced as routine management of AUB in the community.

Literature search
We used the PICO approach to develop a systematic search strategy [33]. We searched MEDLINE via OvidSP, Scopus, and Web of Science. For Medline, key concepts were identified (endometrial hyperplasia/cancer, abnormal uterine bleeding, endometrial sampling), a list of synonyms was generated for each concept and these lists were used to identify MeSH terms for the search (Additional file 1). Similar search strategies were used for Scopus and Web of Science (Additional file 1), always limited to papers from 1984 (when Pipelle was first introduced [34]) to 2016, written in English and involving humans.
We included papers investigating ES in women with AUB. We also considered studies in patients with known cancer; although these studies do not inform the indication of ES in primary care, they were an important source to evaluate test performance. We included review articles and opinion pieces. We excluded papers exclusively analysing postmenopausal patients, papers where the indication was assessment of fertility or recurrent miscarriage and papers where ES was assisted by hysteroscopy (unless this was used as a comparator to blind ES).
The initial search generated 173 results for Medline, 240 for Scopus, and 221 for Web of Science, totalling 634 search hits across all databases, 317 of which were excluded for duplication. The remaining 317 articles were assessed for inclusion using the titles and abstracts. The assessment was independently repeated by a second reviewer and a consensus was reached. After this process, 257 papers were excluded and the full text of 60 papers were read. Twenty-two further papers were excluded while another 22 papers were added from reference search, giving a final list of 60 papers. This selection included 16 randomized controlled trials (RCT), 26 prospective studies, 6 retrospective studies, 5 reviews, 2 meta-analyses, 1 survey, and 4 brief communications and letters to the editor, which were included in the final analysis providing data over 7300 women (Fig. 1).

Bias risk assessment
The quality of the RCTs was assessed using the standard Cochrane Risk of Bias tool [35], and the quality of observational studies was analysed with the modified Agency for Healthcare Research and Quality (AHRQ) quality assessment criteria [36].

Risk of Bias / quality of studies
The overall quality of the RCTs was poor (n = 4) to moderate (n = 12), no high quality studies were identified. For observational studies, the risk of bias ranged from 31 to 79% with a mean weighted score 52.8% SD ± 11.8% which again suggests overall moderate quality [37]. See additional online content for tabulated assessments of individual studies (Additional file 2: TableS1 and 2, Additional file 2).

Five themes
We identified five major themes in the literature: (1) sample adequacy (defined as enough tissue to be analysed by pathologists [38]); (2) test performance when compared with hysterectomy and D&C; (3) acceptability by the patient in terms of pain experienced during sampling; (4) the costs of taking outpatient endometrial biopsies; and (5) the barriers and complications of performing office ES.
All studies, except for one, were carried out in specialised outpatient gynaecology clinics or hospital services (secondary care) [39]. Only one study looked exclusively at premenopausal women [40]. The rest reported on cohorts of both pre-and post-menopausal women or they did not present results based on menopausal status. Most studies included women with symptomatic AUB and no risk of endometrial carcinoma. However, five studies targeted women with endometrial cancer to correlate pre-operative Pipelle with the hysterectomy histopathology [41][42][43][44][45]. Studies are summarised in Table 1.
Five studies including mixed cohorts of pre-and postmenopausal women compared sample adequacy between Pipelle and Tao Brush [30,[57][58][59][60]. Despite one study suggesting that Tao Brush bendable wire should improve sampling of the uterine lateral walls when compared to Pipelle more rigid structure, none of the studies showed significant differences in premenopausal populations [58] (Table 1).
Two studies [40,61] also compared Pipelle to Novak and found no statistically significant difference in terms of adequacy of sample, which varied from 83 to 94% for Pipelle and from 85 to 98% for Novak [40,61] (Table 1). Six additional studies did not find a significant difference when comparing Pipelle with other less popular ES techniques such as Explora [32,62] (85.4-97% for Explora versus 85.2-91% for Pipelle), Tis-u-trap [27] (88% for Pipelle versus 84% for Tis-u-Trap P = 0.5), Endorette Three studies [39,64,65] assessed the ability of Pipelle to retrieve enough tissue for histological analysis without comparing it to other devices, and reported a success rate of 76% in GP practices [39], and a range from 83.01 to 96.82% in secondary care [64,65] (Table 1).
One retrospective study which compared Pipelle samples suggestive of endometrial polyps with subsequent hysteroscopically-guided polypectomies reported Pipelle had a positive predictive value of 55.3% for sampling polyps in premenopausal women [70]. Pipelle has also been reported to have 100% negative predictive value (NPV) for endometrial carcinoma and hyperplasia [24] and up to 99.2% NPV for endometritis and 89.6% for endometrial polyps [52] (Table 1).

Pain / discomfort
Most studies included in this review performed ES on awake patients, but only 23 studies formally assessed patients' pain using visual pain analogue scales and questionnaires (Table 1). A total of 15 studies reported that most patients experienced minimal discomfort with Pipelle [23, 26, 28, 31, 34, 40, 41, 46-49, 54, 61, 71, 72], three did not find any significant difference between Pipelle and Explora [32,62] and Pipelle and Endorette [29], three concluded that Tao Brush was better tolerated than Pipelle [30,57,60] and one study showed less discomfort when using an infant feeding tube as a prototype [63]. A RCT also reported that paracervical lidocaine during Pipelle may decrease pain when compared to placebo [73] (Table 1).

Costs of outpatient endometrial sampling
A total of 17 studies assessed the cost-effectiveness of Pipelle though none formally provided a health economic analysis [23, 24, 26-28, 31, 47-49, 54, 55, 57, 60, 68, 71, 72, 74]. Some of the factors they considered when assessing the total cost of ES were the need for general anaesthesia and hospital admission [23,72] and the cost of operative hysteroscopy/ D&C following a failed office ES or an inadequate sample [55]. Fifteen studies showed Pipelle was cheaper than the alternative ES [23, 24, 26-28, 31, 47-49, 54, 55, 57, 68, 72, 74] and two did not find significant differences between Pipelle and Pipelle Mark 2 [71], and Pipelle and Tao Brush [60]. Two studies concluded that the Vabra was cheaper than Pipelle given its multiple use [26] but when all costs were considered including the need for follow-up for failed procedures, the average cost of Pipelle per patient was approximately 30% cheaper than the Vabra aspirator [55] (Table 1).

Barriers and complications to endometrial sampling in primary care
Several limitations to successful ES were reported including cervical stenosis and pelvic organ prolapse which hindered the access to the uterine cavity [24,75] as well as focal endometrial pathology (e.g. endometrial polyps and submucosal fibroids) and endometrial atrophy which reduces sample adequacy [30,46,69,75,76]. Lack of experience was also linked to inadequate sampling with higher failure rates seen in registrars (39%) than in consultants (25%), (P = 0.13)) [75]. However, a study which compared sample adequacy between nurses (83.3%) and doctors (80%), P > 0.05, concluded that with the right training the ability to perform successful Pipelle is independent of professional category [77].
While few complications have been associated with Pipelle [73]. [78], mainly discomfort and false negative results, a study showed that Pipelle makes the endometrium approximately 0.4 mm thinner and creates echogenic spots which can be misinterpreted as sonographic lesions if the ultrasound is not performed prior to ES [79] (Table 1).

Discussion
Our aim was to search and synthesise the whole range of literature on ES in AUB in low-risk patients to guide further research and develop new evidence-based care pathways in primary care. Overall, the evidence that we have identified supports the use of ES in the outpatient setting and is a valuable source for the development of new care pathways in primary care.
To the best of our knowledge, this study is the first systematic review to primarily focus on the role of ES in assessing and managing AUB in low-risk women in the outpatient setting [25,78,[80][81][82]. The available evidence shows that when Pipelle is combined with clinical assessment and ultrasound findings, it becomes a valuable tool for investigating AUB in low-risk women. Pipelle seems to perform as well or better than any other ES device in terms of sampling adequacy and sensitivity, with comparable results to D&C which for years was the standard technique for obtaining endometrial tissue in patients with AUB [78]. Furthermore, Pipelle seems to be cost-effective and better tolerated in terms of pain/discomfort [83]. However, its use has shown to be limited by cervical stenosis, pelvic organ prolapse and endometrial atrophy [24,75]. Since Pipelle causes changes in the endometrium, it should not be performed before USS [79], and if the ultrasound reports localised lesions, a hospital referral for a hysteroscopy-guided biopsy may prove more useful than performing a blind Pipelle [84] given its limited sensitivity for focal lesions [47,70].
Despite our robust and thorough literature search, we have noted some limitations in the available evidence. We only identified one study which was conducted on a primary care population by general practitioners [39] and one study which looked exclusively at premenopausal patients [40] and therefore, our conclusions are mainly based on studies which were carried out in either outpatient specialised clinics or hospital departments on a mixed cohort of pre-and postmenopausal women. Many of the studies that we identified were of poor or moderate methodological quality with wide-ranging inclusion and exclusion criteria (see Additional file 2). This heterogeneity may partly be responsible for the significant variability seen in terms of the sensitivity and specificity of Pipelle for detecting endometrial hyperplasia/cancer.
A meta-analysis was beyond the scope of this paper but critical appraisal and analysis of pooled data from diagnostic studies is an important next step in establishing the utility of ES. Given the limited information about the true test performance of ES in the community, it is not possible for clinicians to quantify the risk of hyperplasia/cancer (or other pathology) based only on ES. This is especially pertinent when the sample result is normal but the patient is still symptomatic; clinicians should then continue to consider the possibility of false negative results e.g. undiagnosed cancer/hyperplasia in these patients.

Conclusions
The evidence we analysed suggests that performing ES in the outpatient setting may allow effective management of low-risk women with AUB in primary care without referral to a hospital. But the false negative rate, health economics and implications of such a change in practice are still unknown and more research is required.