Endometrial biopsy under direct hysteroscopic visualisation versus blind endometrial sampling for the diagnosis of endometrial hyperplasia and cancer: Systematic review and meta-analysis

Background Endometrial cancer is the most common gynaecological neoplasia in western countries. Diagnosis of endometrial cancer requires an endometrial biopsy. A good quality endometrial biopsy allows not only the identification of the pathology, but also preoperative histologic subtype classification. Endometrial biopsy can be performed under direct hysteroscopic visualisation, but also using blind sampling techniques Objectives To compare endometrial biopsy performed under direct hysteroscopic visualisation versus blind sampling for the diagnosis of endometrial hyperplasia and cancer. Materials and methods Systematic review and meta-analysis. Electronic databases were searched from their inception until March 2022.We included all studies comparing endometrial biopsy performed under direct hysteroscopic visualisation versus blind endometrial sampling. Main outcome measures Sample adequacy, failure rate to detect endometrial cancer or endometrial hyperplasia, and rate of detection of endometrial cancer. The summary measures were reported as relative risk (RR) with 95% of confidence interval (CI). Results Four studies with a total of 1,295 patients were included. Endometrial biopsy under direct hysteroscopic visualisation was associated with a significantly higher rate of sample adequacy (RR 1.13, 95% CI 1.10 to 1.17), and significantly lower risk of failure to detect endometrial cancer or endometrial hyperplasia (RR 0.16, 95% CI 0.03 to 0.92) compared to blind endometrial sampling. However, there was no significant difference between endometrial biopsies taken under direct hysteroscopic visualisation or blindly, with or without a preceding diagnostic hysteroscopy, in the rate of detection of endometrial cancer (RR 0.18, 95% CI 0.03 to 1.06). Conclusion Hysteroscopic endometrial biopsy under direct visualisation is associated with significantly higher rate of sample adequacy and is comparable to blind endometrial sampling for the diagnosis of endometrial cancer and precancer. What is new? Hysteroscopic endometrial biopsy under direct visualisation would be expected to reduce diagnostic failure for endometrial cancer compared to blind endometrial sampling.


Introduction
Endometrial cancer is the most common gynaecological neoplasia in western countries (Randall, 2019). Worldwide, every year more than 350,000 new cases are diagnosed (Ferlay et al., 2019). Endometrial cancer is often diagnosed at an early stage because it frequently causes abnormal vaginal bleeding that prompts timely clinical evaluation (Lu and Broaddus, 2020).
The evaluation of women at risk for endometrial cancer includes transvaginal ultrasound (Jónsdóttir et al., 2021), but the diagnosis requires endometrial biopsy. A good quality endometrial biopsy allows not only the identification of the pathology, but also preoperative histologic subtype classification (Da Cruz Paula et al., 2021). Endometrial biopsy can be performed under direct hysteroscopic visualisation, but also using blind sampling techniques (Di Spiezio Sardo et al., 2020;Papalona et al., 2015;Narice et al., 2018;Rauf et al., 2014). It is unclear whether hysteroscopic biopsy or blind endometrial sampling is superior in detecting significant endometrial disease, endometrial cancer, and endometrial hyperplasia with or without atypia. We therefore conducted a systematic review of the literature to investigate the diagnostic performance of endometrial biopsy performed under direct hysteroscopic visualisation versus blind sampling for diagnosis of endometrial pathology.

Search strategy and selection criteria
This systematic review and meta-analysis were conducted according to a protocol designed a priori and recommended for systematic review (Slim et al., 2003). The meta-analysis was reported following the Preferred Reporting Item for Systematic Reviews and Meta-analyses (PRISMA) statement (Moher et al., 2009). Before data extraction, the review was registered into the PROSPERO International Prospective Register of Systematic Reviews (registration No.: CRD42021245668).
The following electronic databases MEDLINE, Scopus, ClinicalTrials.gov, EMBASE, ScienceDirect, the Cochrane Library at the CENTRAL Register of Controlled Trials, and Scielo were searched from their inception until March 2022. Search terms used were "endometrial cancer", "hysteroscopy", and "biopsy". No restrictions for language or geographical location were applied. In addition, the reference lists of all identified articles were examined to identify studies not captured by electronic searches. The electronic search and the eligibility of the studies were independently assessed by two authors (GS, ADS). Differences were discussed until a consensus was reached.
We included all studies comparing endometrial biopsy performed under direct hysteroscopic visualisation versus blind endometrial sampling for the diagnosis of endometrial cancer or pre-cancerous endometrial pathologies (endometrial hyperplasia with or without atypia). Both observational and randomised trials were included in the review. We planned to include all hysteroscopic settings and all hysteroscopic techniques, e.g., grasp technique, mechanical tissue removal systems or monopolar/ bipolar energy resection. Studies comparing different hysteroscopic techniques but with no blind sampling as a control group, were excluded. The control group included all types of endometrial sampling methods, such as the use of miniature biopsy devices (e.g., Pipelle ® , suction biopsy, Novak curette, vacuum aspiration) and blind dilation and curettage (D&C). We also included studies that used hysteroscopic oriented biopsy in the blind sampling group. Hysteroscopic oriented biopsy was defined as a biopsy performed using a blind technique immediately after a diagnostic hysteroscopy. Studies comparing different blind techniques, e.g., Pipelle ® vs D&C, with no hysteroscopic approach as intervention group were excluded. Case reports and studies including less than 5 patients were excluded.

Data extraction and risk of bias assessment
Two reviewers (ADS, GS) independently assessed the risk of bias of the included studies via the Methodological Index for Non-Randomized Studies (MINORS) (Slim et al., 2003). Seven domains related to risk of bias were assessed in each study: 1) Aim (clearly stated aim), 2) Rate (inclusion of consecutive patients and response rate), 3) Data (prospective collection of data), 4) Bias (unbiased assessment of study endpoints), 5) Time (follow-up time appropriate), 6) Loss (loss to follow-up), 7) Size (calculation of the sample size). Review authors' judgments were categorised as "low risk," "high risk" or "unclear risk of bias." Discrepancies were resolved by discussion with a third reviewer (BZ). Additional data were asked from the authors of the original studies, if feasible.

Primary and secondary outcomes
All analyses were done using an intention-totreat approach, evaluating women according to the treatment group to which they were randomly allocated in the original study. The primary outcome was sample adequacy, defined as enough tissue quantity and quality to be analysed by pathologists. The secondary outcomes were failure to detect endometrial cancer or endometrial hyperplasia (McCluggage, 2006), and mean procedure length for sampling.

Statistical analysis
The data analyses were completed using Review Manager v. 5.3 (The Nordic Cochrane Centre, Cochrane Collaboration, 2014, Copenhagen, Denmark). The summary measures were reported as summary relative risk (RR) or as summary mean difference with 95% of confidence interval (CI) using the fixed effects model. I-squared (Higgins I2) greater than 0% was used to identify heterogeneity. Data from each eligible study were extracted without modification of original data onto custom-made data collection forms. 2 by 2 contingency tables were constructed and relative risks (RR) calculated. For continuous outcomes means ± standard deviation (SD) was extracted and imported into Review Manager. Potential publication biases were assessed statistically by using Begg and Egger's tests. A p value <0.05 was considered
Therefore, 4 studies (Ceci et al., 2002;Wanderley et al., 2016;Rosenblatt et al., 2017;Di Spiezio Sardo et al., 2020) with a total of 1,295 participants, were included in the meta-analysis. Publication bias was assessed statistically by using Begg's and Egger's tests, showed no significant bias (P=0.69 and P=0.51, respectively). The quality of the studies included in our meta-analysis is reported in was associated with significantly higher rate of sample adequacy (RR 1.13, 95% CI 1.10 to 1.17; Figure 3), although there was considerable statistical heterogeneity (I2=97%). There was a significantly lower risk of failure to detect endometrial cancer or endometrial hyperplasia (RR 0.16, 95% CI 0.03 to 0.92; I2=0%; Figure 4) compared to blind sampling. There was no significant difference between endometrial biopsies taken under direct hysteroscopic visualisation or blindly, with or without a preceding diagnostic hysteroscopy, in the rate of detection of endometrial cancer (RR 0.18, 95% CI 0.03 to 1.06: Figure 5). Whilst the point estimate for detection of endometrial cancer favoured direct hysteroscopic biopsy, the data were derived from two studies only and statistical significance was not reached ( Figure 5). No differences were found in the mean procedure length for sampling (44± vs 47±38 seconds; mean difference -3.00 seconds, 95% CI -35.91 to 29.91).

Main findings
This systematic review aimed to compare sample adequacy and failure rates of endometrial biopsy performed under direct hysteroscopic visualisation versus blind endometrial sampling for the diagnosis of endometrial hyperplasia and cancer. Four studies (Ceci et al., 2002;Wanderley et al., 2016;Rosenblatt et al., 2017;Di Spiezio Sardo et al., 2020), with a total of 1,295 participants, were included in the meta-analysis. Endometrial biopsy under direct hysteroscopic visualisation was associated with significantly higher rate of sample adequacy compared to blind sampling.  Table I shows the characteristics of the included studies. All studies used hysterectomy as the diagnostic reference standard except for Wanderley et al. (2016) where the reference standard was not reported. The indications for hysterectomy were suspected cancer in one study (Di Spiezio Sardo et al., 2020); abnormal bleeding, polyps or a postmenopausal endometrial thickness (ET) >4mm in one study (Ceci et al., 2002); abnormal bleeding, or a postmenopausal ET >4mm or premenopausal ET >15mm in one study (Wanderley et al., 2016); while the indication was not reported in one study (Rosenblatt et al., 2017). All studies included women of pre-and postmenopausal status, apart from Rosenblatt et al. (2017) which restricted recruitment to postmenopausal women only.
It should be noted that the retrospective study by Ceci et al. (2002) included 443 patients who underwent office hysteroscopy followed by hysterectomy. The results of this study were then compared with a historical control of a previous study in which the same group of researchers examined the diagnostic accuracy of dilatation and curettage (D&C) with hysterectomy as the diagnostic reference standard (Bettocchi et al., 2001). Figure 4 show the forest plots for primary and secondary outcomes. Endometrial biopsy under direct hysteroscopic visualisation    Hysteroscopic visualisation was also associated with 82% decreased risk of failure to detect endometrial cancer, although statistical significance was not reached (p=0.06). Pooled data did not report any significant differences in the mean procedure length for sampling between the two techniques, with a mean of about 44-47 seconds.

Strengths and Limitations
We conducted a comprehensive search and followed standard approaches to conducting a systematic quantitative review (Cumpston et al., 2019).
cancer is usually transvaginal ultrasound, followed by endometrial biopsy. A good quality endometrial biopsy allows not only the diagnosis of endometrial cancer but also the histologic subtype classification. Currently, there is a variety of endometrial sampling methods, including blind sampling with Pipelle ® , blind D&C, hysteroscopy-oriented biopsy, or hysteroscopic endometrial biopsy under direct visualisation. Diagnostic accuracy studies of endometrial cancer showed high diagnostic accuracy when the endometrial biopsy is obtained under direct hysteroscopic visualisation (Clark et al., 2002), and low to moderate when collected by blind D&C (Bettocchi et al., 2001;Vorgias et al., 2003). A large number of papers have extensively shown throughout the years the significant limitations of the blind technique, including the need for in-patient admission and general or regional anaesthesia; the high risk of complications; poor diagnostic accuracy (high number of focal lesions missed); and the total absence of any therapeutic role (Bettocchi et al., 2001).
However, findings from this systematic review and meta-analysis are limited by the observational nonrandomised study design of the studies included. Of the four studies that were included in the final analysis only one had a prospective study design (Rosenblatt et al., 2017). The source studies were heterogeneous, limiting the ability to draw meaningful conclusions from the pooled analyses. The main limitation of the review was the low quality of the included studies.
In particular, one of the included (Wanderley et al., 2016) studies did not report the reference standard used to evaluate the methods of endometrial sampling against. Considering the methodological deficiencies of the primary studies we were unable to construct 2x2 contingency tables to assess overall diagnostic accuracy.

Implication
Endometrial carcinoma is the most common gynaecological cancer in western countries. After history taking and physical examination, the first step in the workup of a patient with suspected endometrial    However, despite this evidence, the Society of Gynecologic Oncology and the American Congress of Obstetricians and Gynecologists still emphasise the diagnostic and therapeutic role of D&C (Practice Bulletin No 149, 2015). Notably, when dealing with endometrial cancer, it is important to distinguish between diffuse or focal cancer (Patel et al., 2010). Indeed, the value of any blind procedure is when it reports a positive result, when it is negative (especially in cases of focal pathology or early adenocarcinoma) it can be a false negative and therefore hysteroscopy may be required (Clark, 2017;van Hanegem, 2017). It is possible that failure to adopt hysteroscopically directed endometrial biopsy reflect the need to take multiple samples requiring several instrument insertions due to the small amount of tissue obtained with conventional 5Fr / 7Fr forceps. However, with the introduction of mechanical hysteroscopic tissue removal (mHTR) systems, large, representative endometrial tissue samples can easily be obtained without the need for repeated reinsertion of the hysteroscope (Franchini 2021). Robust, diagnostic accuracy studies are needed to compare the accuracy of mHTR against blind endometrial sampling and / or conventional hysteroscopic sampling methods.

Conclusion
In summary, hysteroscopic endometrial biopsy under direct visualisation is associated with a significantly higher rate of sample adequacy and is comparable to blind endometrial sampling for the diagnosis of endometrial cancer and precancer. A large, well-designed, randomised controlled trial, is needed to confirm our findings.