Sage Journals: Discover world-class research

Abstract

Postmenopausal bleeding is associated with an elevated risk of having endometrial cancer. The aim of this review is to give an overview of existing prediction models on endometrial cancer in women with postmenopausal bleeding. In a systematic search of the literature, we identified nine prognostic studies, of which we assessed the quality, the different phases of development and their performance. From these data, we identified the most important predictor variables. None of the detected models completed external validation or impact analysis. Models including power Doppler showed best performance in internal validation, but Doppler in general gynecological practice is not easily accessible. We can conclude that we have indications that the first step in the approach of women with postmenopausal bleeding should be to distinguish between women with low risk versus high risk of having endometrial carcinoma and the next step would be to refer patients for further (invasive) testing.

Keywords

endometrial carcinoma multivariate analysis postmenopausal bleeding prediction models prognostic studies

Endometrial carcinoma is the most common gynecologic malignancy. Approximately 95% of women with endometrial carcinoma present with postmenopausal bleeding (PMB) [1,2]. PMB signals endometrial carcinoma, which is present in approximately 10% of cases [3,4], or less serious conditions, such as benign endometrial polyps or endometrial atrophy [3,5 –7].

To reduce invasive procedures in women with PMB, measurement of the endometrial thickness is used to stratify women into low versus high risk of having endometrial carcinoma. Measurement of endometrial thickness has shown to be accurate in excluding endometrial cancer, although the risk of endometrial carcinoma with a negative test is still 0.7–3.5% depending on the cut-off point used [8,9].

In women with PMB there is considerable variability in endometrial thickness and the likelihood of endometrial carcinoma [10]. Individual patient characteristics, including age, time since menopause, obesity, hypertension, diabetes mellitus and reproductive factors, are associated with a higher risk of endometrial carcinoma [10 –16]. While the probability of PMB decreases with increasing age [17], the probability of endometrial cancer in women with PMB increases significantly with increasing age. The probability rises from 1% in women younger than 50 years of age to 24% in women older than 80 years of age [18].

In clinical practice, tests are commonly combined in diagnostic sequences and disease probabilities are usually estimated in a hierarchical manner: first combining information from history and examination, followed by additional information obtained from diagnostic tests. The post-test probability is not only dependent on test characteristics, but also on the pretest probability, which is altered by patient characteristics. However, current diagnostic policy in women with PMB is not based on these patient-specific risk factors, but only on one fixed cut-off point for endometrial thickness [2,19 –21].

Clinical doctors want to identify women with a high risk for endometrial cancer when presenting with PMB. Several articles have studied this subject and developed models to estimate the individual chance of endometrial carcinoma in women presenting with PMB. The purpose of this review is to give an overview of the existing prediction models for endometrial carcinoma in women with PMB, to assess their quality and to identify important predictor variables.

Methods

Study identification

We performed a computerized MEDLINE and EMBASE search to identify all studies on prediction models in women with PMB published from inception to June 2011. The search was limited to human studies, no restrictions were held concerning publication year or language. We included articles reporting on multivariable models predicting endometrial cancer in women with PMB. We checked references cited in the selected articles for further relevant prediction models not identified by the electronic searches. We used all known synonyms for the terms ‘PMB’ and ‘endometrial cancer’ and we used a search filter for prediction models [22]. The search strategy can be found in the Appendix.

Study selection

This review focused on articles that report on a prediction model for endometrial carcinoma in women with PMB. In this review, a prediction model was defined as a multivariable model that expresses the chance of endometrial carcinoma as a function of two or more predictor variables. PMB was defined as vaginal bleeding after more than 1 year of amenorrhea after the age of 40 years, or persistent (>3 months) unscheduled bleeding on HRT.

Two independently working reviewers (N van Hanegem and MC Breijer) selected the articles by assessing titles and abstracts. If there were any doubts about eligibility after reading the title and abstract, we read the full text version to make sure no articles were missed. In case of a disagreement, the article was included for full text reading and assessed by a third reviewer (A Timmermans).

Study quality assessment

A framework for quality assessment was developed based on the recommendations of Hayden et al. [23] and on a quality assessment framework for prediction models in subfertile women to predict the chance of pregnancy [24]. The framework was divided into four sections: study participation, predictor variables, outcome measurement and analysis. Each item in the different sections was scored with yes, no or unclear.

Predictor variables

All predictor variables were collected for each prediction model. The predictor variables are the potential predictors, which were tested both during model development and in the final model. The original articles selected multiple variables or risk factors that are thought to be associated with an increased risk of endometrial cancer. These variables have been tested in the original articles for univariate association and, if sufficiently contributing to predictive accuracy in multivariable regression analysis, combined to construct a clinical prediction model. We collected all different predictor variables from the original articles, together with their significance, to identify the most important predictor variables for endometrial cancer. The most important predictor variables had been considered as statistically significant input variables in three or more studies or were considered statistically significant in two studies and had not been tested in other studies.

Model development assessment

The development of a prediction model consists of three phases: model derivation, model validation and impact analysis [25]. In the first phase, model derivation, predictor variables are identified by logistic regression. Model validation, the second phase, consists of an internal and external validation phase [24]. In internally validated models, the performance of the model is tested in the same data set in which the model was developed, or in a group of subsequent patients within the same center. In external validation, the goal is to demonstrate generalizability and reproducibility in patients different from the patients used for derivation of the original model. Therefore, the prediction model is evaluated on new data collected from an appropriate patient population in a different center [26]. The final phase of model development is impact analysis, in which prediction models are tested for their ability to change clinicians' decisions and to change patient outcomes [27]. All prediction models identified in this review are classified into the different phases of model development. We sent an email to all authors of the identified articles to investigate if their models are undergoing external validation and are not published yet.

Model performance

Performance measures (calibration, discrimination and clinical usefulness) and the range of probabilities given by the different prediction models were recorded. Calibration refers to the agreement between observed probabilities and predicted probabilities for groups of patients; this is usually reported as a calibration plot or a Hosmer–Lemeshow statistic (test for ‘goodness-of-fit’) [28]. Discrimination is commonly reported as the c-statistic (concordance), also referred to as the area under the receiver-operating characteristic curve (AUC). It measures the ability of a prediction model in separating patients with endometrial cancer and patients without endometrial cancer. An AUC of 0.5 describes a non-informative test, whereas an AUC of 1.0 represents a test that discriminates perfectly between presence and absence of a disease [29]. Clinical usefulness measures how close a prediction for an individual patient is to their actual outcome. This is mostly reported as accuracy (percentage of patients correctly classified), sensitivity or specificity, positive or negative predictive value (NPV), or likelihood ratios (LR) of a prediction model [30]. As we are interested in identifying a group of patients with a high risk for endometrial cancer, we are most interested in a high sensitivity, high NPV and a low negative LR.

Results

Study identification & selection

Of 754 articles identified by the MEDLINE and EMBASE search, a total of nine articles met the inclusion criteria of our review [31 –39]. We identified another three articles by scanning the reference lists of included articles [40 –42]; however, none of these matched our inclusion criteria after reading the abstract and full text version of these articles (Figure 1).

Figure 1.

Study selection diagram.

Study characteristics

Study characteristics are shown in Table 1. Of the nine selected articles on prediction models for women with PMB, five articles described the development of one model and four articles described two or more different prediction models. In the nine selected articles, four models were based primarily on patient characteristics [33,35,38,39], four prediction models were based on a combination of patient characteristics and gray-scale TVS findings [33 –35,37], two prediction models were based on a combination of patient characteristics, hysteroscopy and/or gray-scale TVS findings [33], two prediction models were based on TVS findings only [32,36] and three models used Doppler TVS findings as a predictor variable [31,36,39].

Table 1.

Study characteristics of included articles.

Study (year)	Patients	Exclusion criteria	Individuals	Malignancies, n (%)	Age (years)	Study design	Outcome	Prediction model	Ref.
Epstein et al. (2002)	PMP bleeding, ET ≥5 mm	Doppler artefacts, incorrect processing	83	16(19)	Mean 66	Prospective CH	Endometrial carcinoma	Two models: both combination patient history, TVS + Doppler	[31]
Randelzhofer et al. (2002)	PMP bleeding	HRT/tamoxifen use	321	95(30)	47–91 (median 68)	Prospective CH	Endometrial carcinoma	One model: TVS only	[32]
Bachmann et al. (2003)	PMP bleeding, unscheduled bleeding on HRT	–	428	19(4.6)	28–61 (median 54)	Prospective CH	Endometrial carcinoma	Four models: patient history, patient history + TVS, patient history + hyst, patient history + TVS + hyst	[33]
Bruchim et al. (2004)	PMP bleeding	HRT use	95	9(9.5)	46–94 (mean 60)	CH†	Endometrial carcinoma	One model: patient history + TVS	[34]
Opmeer et al. (2007)	PMP bleeding	History of hysterectomy, HRT use, recurrent bleeding, TVS not possible/not performed	540	56(10.3)	37–91 (mean 62)	Prospective CH	Endometrial carcinoma or pre-malignancy	Two models: one patient characteristics, one combined model with patient characteristics + TVS	[35]
Opolskiene et al. (2007)	PMP bleeding, ET ≥4.5 mm	Fluid in cavity, absence of power Doppler signals, large myomas, no histologic diagnosis	120	30 (25)	†	Prospective CH	Endometrial carcinoma	Many models: TVS + Doppler	[36]
Burbos et al. (2010)	PMP bleeding	Asymptomatic women	3047	–	35–97 (median 59)	Prospective CH	Endometrial carcinoma	One model: patient history + TVS	[37]
Burbos et al. (2011)	PMP bleeding	History of hysterectomy, asymptomatic women	3548	201 (6)	54–73	Prospective CH	Endometrial carcinoma	One model: patient history only	[38]
Opolskiene et al. (2011)	PMP bleeding, ET ≥4.5 mm	Absence of processed ultrasound images or reliable histologic diagnosis	261	63 (24)	47–91 (median 67)	Prospective CH	Endometrial carcinoma	Four models: one based on patient history, one patient history + TVS, two patient history + TVS + Doppler	[39]

†

Not well described.

CH: Cohort study; ET: Endometrial thickness; Hyst: Hysteroscopy (positive/negative); PMP: Postmenopausal; TVS: Transvaginal sonography.

Patient selection and inclusion criteria were not the same in all articles. All nine articles included women with PMB, but three of these articles studied a population of women with a high-risk profile for endometrial cancer, based on an endometrial thickness of ≥5 mm [31,36,39].

Study quality

The results of the quality assessment are reported in Figure 2. Overall, study quality was good. The quality of the description of the setting and study period was rated as moderate; this was not described in three out of nine articles. Three articles included all women with PMB, but performed histology only in patients with an increased endometrial thickness. All three articles explained that no further investigations were performed in women with an endometrial thickness <5 mm, because evidence suggests a very low probability of cancer below this threshold [35,37,38].

Figure 2.

Quality of included studies.

Predictor variables

The nine included articles investigated 27 different possible prediction variables (Table 2) . Age was tested in all nine articles and turned out to be statistically significant in multivariable analysis in six articles and was used in the prediction model in six articles. Endometrial thickness was tested in eight articles, statistically significant in multivariable analysis in eight articles and used in eight prediction models. Most important predictor variables in patient history were: age, BMI, diabetes, frequency of bleeding, use of anticoagulants and HRT. Endometrial thickness, endometrial morphology and endometrial border were identified as significant gray-scale TVS variables. In the three articles studying the use of Doppler for predicting endometrial cancer, endometrial color score and vascularity index were identified as the most important predictor variables.

Table 2.

Predictor variables evaluated and used in the prediction models.

	Epstein et al. (2002) [31]	Randelzhofer et al. (2002) [32]	Bachmann et al. (2003) [33]	Bruchim et al. (2004) [34]	Opmeer et al. (2007) [35]	Opolskiene et al. (2007) [36]	Burbos et al. (2010) [37]	Burbos et al. (2011) [38]	Opolskiene et al. (2011) [39]
Patient history variables
Age	3	2	1	1	1	2	1	1	1
BMI					1		1	1	3
Diabetes			3		1		1	1	3
FOB							1	1
Amount of bleeding							3	3
HRT use	1		1			3	3	3	1
Anticoagulant use					1				1
Nulliparity					1				3
TM	3				2
VAS score									1
Examiner	3
Hypertension			3		2		2	2	3
Thyroid dysfunction				3				3
History of malignancy					3		2	2
Ultrasound variables: gray-scale
ET	1	1	1	1	1	1	1		1
Heterogeneous echogenicity/morphology	1	1				1
Endometrial border		1				1
Intracavitary fluid		3
Endometrial area	2
Ultrasound variables: Doppler
Vascularized area	2					1
Endometrial color score	1								1
Irregular branching						1
Vascularity index	1						1
MIP in endometrial area	2
MIP in vascularized area	3
Hysteroscopy variables
Suspicious hysteroscopy findings		1

Statistically significant in multivariate analysis and included in prediction model.

Statistically significant in univariate analysis and not included in prediction model.

Not statistically significant and not included in prediction model.

ET: Endometrial thickness; FOB: Frequency of bleeding; MIP: Mean intensity of pixels; TM: Time since menopause; VAS: Visual analog scale.

Phases of model development

All articles selected in this review addressed the first phase of developing a prediction model: model derivation [24]. Of the nine articles on predicting endometrial cancer in women with PMB, eight had been internally validated but none of these models passed the external validation phase. We asked all six research groups, who developed the nine different prediction models, if their models are undergoing external validation and we received response from all of them. The two prediction models of Opolskiene et al. [36,39] are undergoing temporal validation (internal validation in a newly recruited patient group) and external validation in an international multicenter study by Valentin et al. [Valentin L, Sladkevicius P, Pers. Comm.]. No results are available yet, since they are still recruiting patients for these studies. The two prediction models developed by Burbos et al. [37,38] were recently used in an article to compare the performance in internal validation of these models [43]. This group is working on external validation. Finally, we can report that the prediction model of Opmeer et al. is currently being externally validated in two cohorts [35]: one cohort in three different hospitals in The Netherlands and one in Skåne University Hospital Malmö (Scania, Switzerland) in collaboration with the group of Valentin et al. [Valentin L, Sladkevicius P, Pers. Comm.], but this external validation is not published yet. There were no impact analysis studies (i.e., studies that showed that the prediction model indeed improved patient outcome or was cost-effective in clinical practice).

Performance of the prediction models

The performance of the eight articles that internally validated their models [31 –33,35 –39] is presented in Table 3.

Table 3.

Evaluation of model development and model performance.

Study (year)	Development phase	Prediction model	Specification of model performance at internal validation				Ref.
			Discrimination†	Calibration	Clinical usefulness	Result as reported in paper
Epstein et al. (2002)	Internal validation	Subjective prob of malignancy (PH, US, Doppler)	0.88	–	Sens: 0.75; spec: 0.96	Power Doppler can contribute to diagnosis of endometrial cancer	[31]
		Objective prob of malignancy (PH, US, Doppler)	0.88	–	Sens: 0.88; spec: 0.81
Randelzhofer et al. (2002)	Internal validation	ET + endometrial structure + myometrial border (cut-off point ET >10 mm)	–	Estimated prob 2.8 vs 2.1% real malignant	Sens: 0.97; spec: 0.62; NPV: 0.98; accuracy: 0.72	The combined assessment of ET and endometrial morphology may improve diagnostic accuracy	[32]
Bachmann et al. (2003)	Internal validation	PH	0.8	–	–	Not much increased value in testing with US if hysteroscopy was already performed	[33]
		PH + ET	0.82
		PH + hysteroscopy	0.910
		PH + ET + hysterscopy	0.914
Bruchim et al. (2004)	Model derivation	ET + time since menopause	–	–	–	Time since menopause and ET can define when invasive testing is needed	[34]
Opmeer et al. (2007)	Internal validation	PH	0.76	–	NPV: 0.990; efficiency: +0.6	Compared with US only, efficiency gain is reflected in increased AUC and reduced number of procedures, with PH + US in a sequential strategy	[35]
		PH + US if prob>4%	0.76		NPV: 0.990; efficiency: −0.16
		PH + US, histology if prob >4%	0.9		NPV: 0.996; efficiency: −0.05
Opolskiene et al. (2007)‡	Internal validation	ET, echogenicity (and Doppler)	0.91/0.92§	–	Sens: 0.93/0.87§; spec: 0.79/0.83; LR-: 0.1/0.2	A model including ET and heterogeneous echogenicity of the endometrium was best in predicting endometrial carcinoma, with Doppler diagnostic performance improved marginally	[36]
Burbos et al. (2010)	Internal validation	DEFAB ≥3	0.77/0.66	–	Sens: 0.82; spec: 0.50; accuracy: 0.52; LR-: 0.36	Fair accuracy in separating women without cancer from women with cancer	[37]
Burbos et al. (2011)	Internal validation	FAD 31 (cut-off FAD 31 ≥4)	0.73/0.66	–	Sens: 0.80; spec: 0.51; accuracy: 0.53; LR-: 0.39	Reasonable discriminatory ability	[38]
Opolskiene et al. (2011)	Internal validation	PH	0.74	–	Sens: 0.89; LR-: 0.22	Fairly good for excluding endometrial cancer when power Doppler is added	[39]
		PH + ET	0.82	–	Sens: 0.84; LR-: 0.24
		PH + ET + Doppler (VAS)	0.89	–	Sens: 0.70; LR-: 0.31
		PH + ET + Doppler (VI)	0.91	–	Sens: 0.89; LR-: 0.14

†

Discrimination is reported as AUC.

‡

As many models are described, we selected the model with the best performance.

Including Doppler.

AUC: Area under the receiver-operating characteristic curve; DEFAB: Diabetes, endometrial thickness, frequency of bleeding, age and BMI; ET: Endometrial thickness; FAD 31: Frequency of bleeding, age, diabetes, BMI cut-off 31; LR-: Negative likelihood ratio; NPV: Negative predictive value; PH: Patient history; Prob: Probability; Sens: Sensitivity; Spec: Specificity; US: Ultrasound; VAS: Visual analog scale; VI: Vascularity index.

Calibration was described in one article [32]. The estimated probability of malignancy and the observed proportion of patients with endometrial carcinoma are mentioned in Randelzhofer et al.'s article [32]. However, calibration is generally reported as a calibration plot. None of the studies reported on calibration in a calibration plot.

Discrimination was studied in seven out of eight articles by calculating an AUC. The AUC varied from 0.66 to 0.92 for different prediction models, with the highest AUC for a model combining Doppler and gray-scale transvaginal sonography (TVS) [36].

In all internally validated studies clinical usefulness is described, with the highest sensitivity and the lowest negative LR for a combined model with patient characteristics, gray-scale TVS and Doppler [39]. The highest NPV found for a model was 0.996, which combined patient history, endometrial thickness and histology in a sequential strategy [35]. The performance of the four models using only patient characteristics showed a high sensitivity or high NPV in two models [35,39] and a low LR for a negative outcome in one model [39].

All three studies in which Doppler was studied as a predictor variable, reported this information to contribute to the prediction of endometrial carcinoma in women with PMB [31,36,39]. Endometrial thickness was used as a variable in eight prediction models and seven found that incorporating endometrial thickness may improve diagnostic accuracy of a model.

Discussion

We systematically reviewed existing prediction models for endometrial carcinoma in women with PMB and to identify the most important predictor variables. We found nine studies reporting on the development of prediction models for endometrial carcinoma in women with PMB. Eight of these studies described at least one aspect of internal validation and, until now, none of the prediction models have been externally validated.

The different predictor variables can roughly be divided into four subjects: patient characteristics, gray-scale TVS variables, Doppler TVS variables and hysteroscopy variables. Most prediction models used a combination of these subjects to predict the chance of endometrial carcinoma. We chose to limit our list of the most important predictor variables to those that had been considered as statistically significant input variables in three or more studies and to those that were significant input variables in two studies and had not been tested in other studies. By doing this, we identified the most important variables without missing possible important variables that have not yet been extensively studied. Using these limits we identified 11 important input variables for predicting endometrial cancer in women with PMB (Table 2).

Almost all articles reported performance in terms of discrimination and/or clinical usefulness, whereas calibration was reported only incidentally. In this study, we identified five articles describing a prediction model with good discrimination (AUC: >0.8) [31,33,35,36,39]. As only one study described data on calibration, there is insufficient data available to draw conclusions on calibration.

Two studies showed best performance regarding discrimination and clinical usefulness: Opolskiene et al. and Opmeer et al. [35,39]. In the model by Opolskiene et al., a combination of patient characteristics, gray-scale TVS and Doppler was used. They concluded that their model excludes endometrial cancer reasonably well when power Doppler is added. Furthermore, in all three studies that used Doppler, Doppler was found to contribute to the prediction of endometrial carcinoma in women with PMB [31,36,39]. Based on this, we could conclude that the best model in predicting endometrial cancer is a model that uses a combination of patient characteristics, endometrial thickness and power Doppler. However, power Doppler cannot be used in all patients. All three Doppler models excluded patients based on different reasons: Doppler artifacts, incorrect processing of TVS image, fluid in the cavity and absence of Doppler signals or large myomas. Another limitation in the use of power Doppler is that these studies do not give information on the interobserver variability and learning curve in measuring Doppler variables. For application of results found in Doppler studies it is important to use the same ultrasound system, as the color content of a power-Doppler scan depends heavily on Doppler sensitivity [39].

Although the performance of the models using Doppler seems reasonable, a model using patient characteristics and endometrial thickness may be more useful in daily practice. In a healthcare system with general practitioners referring patients with a high risk of malignant disease to a specialist, the best model would be a model that can distinguish women with a high risk of endometrial cancer from women with a low risk based on patient characteristics only. Such a model would also be useful in situations where TVS is not directly available. Only women with a high risk could be referred for TVS or to the gynecologist for further evaluation and women with a low risk could be reassured and referred only at recurrent bleeding. Based on this review we could not identify a model with a good performance in internal validation based on patient characteristics only. However, two of four models based on patient characteristics showed good performance in clinical usefulness with a high sensitivity, a high NPV and/or a low LR for a negative outcome [35,39]. Based on these results we can conclude that although these models do not show a high AUC, they could be useful in clinical practice. These models were found to discriminate women with a high risk for endometrial cancer from women with low risk and to select women for further (invasive) testing.

The above conclusions are based on reported model performance based on internal validation only. To implement a prediction model into clinical practice, external validation is essential. McGinn et al. describe three reasons [25]. First, a prediction model may reflect associations between given predictors and outcomes that are primarily due to chance. Second, the predictor variables used in a model may be idiosyncratic to that specific population, which suggests that the prediction model may fail in a new setting. Third, clinicians may fail to implement the model comprehensively or accurately in their clinical practice. The result would be that a model succeeds in theory, but fails in practice. For a successful implementation, a model should be validated both internally and externally and finally go through the phase of impact analysis in the same population from which a model is derived. As none of the prediction models have completed the phase of external validation, they cannot be used in clinical practice yet.

When evaluating these prediction models by external validation or finally in impact analysis, one should keep in mind that these models were developed in different patient populations. The target population in which a model is derived should be the same as the population in which a model is tested or clinically used. Selecting a high-risk population (e.g., a population with an endometrial thickness of ≥5 mm) will result in a different performance and possibly in the selection of different predictor variables compared with an unselected population of women with PMB. Furthermore, implicit selection mechanism could occur within a population, for example within a general population or a population within a gynecological practice, or within health systems in different countries. Different populations have different prevalence of endometrial cancer, which could be an explanation for the differences found in the performance of the models.

A consensus has not been found in systematic reviews or in international guidelines regarding the best sequence of diagnostic procedures for women with PMB [8]. Considering the performance of the existing prediction models, we can conclude that we have indications that the first step in the approach of women with PMB should be to distinguish between women with low versus increased risk of having endometrial carcinoma, and the next step would be to refer patients for TVS or further invasive testing.

Future perspective

The prediction models that have been developed for women with PMB showed good performance, but have only reached the phase of internal validation. Future research should focus on external validation and impact analysis of these prediction models. We hope that these will confirm their prognostic abilities, so that in the next few years prediction models can be implemented in general gynecological practice. Based on this review, we conclude that clinical prediction models show promising results, but further external validation is required as well as impact analysis to maximize diagnostic accuracy of the models at an acceptable patient burden and for acceptable healthcare costs.

Executive summary

Background

Postmenopausal bleeding (PMB) is associated with an elevated risk of having endometrial cancer. Clinical and ultrasound characteristics influence this risk in the individual patient.

Methods

We systematically reviewed nine prediction models for women with PMB.

Results

The most important predictor variables in women with PMB are: age, BMI, diabetes, frequency of bleeding and use of anticoagulants and HRT (patient characteristics), endometrial thickness, endometrial color score and endometrial border (gray-scale ultrasound) and endometrial color score and vascularity index (Doppler).

Models including power Doppler showed the best performance in internal validation, but based on the difficult use of Doppler in general gynecological practice, we concluded that the best models up to present, are models combining patient characteristics with endometrial thickness.

Discussion

Eight models were internally validated, with best performance in a study combining patient characteristics and measurement of endometrial thickness. Doppler is found to contribute to the prediction of endometrial carcinoma, but cannot be used in all patients.

The first step in the approach of women with PMB should be to distinguish between low- and high-risk patients for having endometrial carcinoma and the next step would be to refer patients for transvaginal sonography or further invasive testing.

Footnotes

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.

Appendix

References

Papers of special note have been highlighted as:.

of interest.

of considerable interest.

Brenner

. Differential diagnosis of abnormal uterine bleeding. Am. J. Obstet. Gynecol. 175 (3 Pt 2), 766–769 (1996).

Dutch Society of Obstetrics and Gynecology (NVOG). Diagnostics in Abnormal Vaginal Blood Loss in the Postmenopausal Period. Dutch Society of Obstetrics and Gynecology, Utrech, The Netherlands (2003).

Dijkhuizen

Brolmann

Potters

The accuracy of transvaginal ultrasonography in the diagnosis of endometrial abnormalities. Obstet. Gynecol. 87(3), 345–349 (1996).

Emanuel

Verdel

Wamsteker

Lammes

. An audit of true prevalence of intrauterine pathology: the hysteroscopic findings, controlled for patient selection in 1202 patients with abnormal uterine bleeding. Gynecol. Endosc. 4, 237–241 (1995).

Timmermans

AVS

. Office hysteroscopy in women with postmenopausal bleeding: see and treat of endometrial polyps using a Duckbill polyp snare. Gynaecol. Surg. 12(1), 189–190 (2004).

Epstein

Ramirez

Skoog

Dilatation and curettage fails to detect most focal lesions in the uterine cavity in women with postmenopausal bleeding. Acta Obstet. Gynecol. Scand. 80(12), 1131–1136 (2001).

10.

Ferrazzi

Torri

Trio

Sonographic endometrial thickness: a useful test to predict atrophy in patients with postmenopausal bleeding. An Italian multicenter study. Ultrasound Obstet. Gynecol. 7(5), 315–321 (1996).

11.

van Hanegem

Breijer

Khan

Diagnostic evaluation of the endometrium in postmenopausal bleeding: an evidence-based approach. Maturitas 68(2), 155–164 (2011).

12.

Systematic review on the diagnostic pathway in women with postmenopausal bleeding .

13.

Breijer

Timmermans

van Doom

Diagnostic strategies for postmenopausal bleeding. Obstet. Gynecol. Int. 2010, 850812 (2010).

14.

van Doom

Dijkhuizen

Kruitwagen

Accuracy of transvaginal ultrasonography in diabetic or obese women with postmenopausal bleeding. Obstet. Gynecol. 104(3), 571–578 (2004).

15.

Anderson

Mink

Diabetes and endometrial cancer in the Iowa Women's Health Study. Cancer Epidemiol. Biomarkers Prev. 10(6), 611–616 (2001).

16.

Gull

Karlsson

Milsom

Factors associated with endometrial thickness and uterine size in a random sample of postmenopausal women. Am. J. Obstet. Gynecol. 185(2), 386–391 (2001).

17.

Weiderpass

Persson

Adami

Body size in different periods of life, diabetes mellitus, hypertension, and risk of postmenopausal endometrial cancer (Sweden). Cancer Causes Control 11(2), 185–192 (2000).

18.

Xiang

Ruan

Menstrual and reproductive factors and endometrial cancer risk: results from a population-based case—control study in urban Shanghai. Int. J. Cancer 108(4), 613–619 (2004).

19.

McPherson

Sellers

Potter

Reproductive factors and risk of endometrial cancer. The Iowa Women's Health Study. Am. J. Epidemiol. 143(12), 1195–1202 (1996).

20.

Reeves

Pirie

Beral

Cancer incidence and mortality in relation to body mass index in the million women study: cohort study. BMJ 335 (7630), 1134 (2007).

21.

Astrup

Nde

Olivarius F

. Frequency of spontaneously occurring postmenopausal bleeding in the general population. Acta Obstet. Gynecol. Scand. 83(2), 203–207 (2004).

22.

Gredmark

Kvint

Havel

Histopathological findings in women with postmenopausal bleeding. Br. J. Obstet. Gynaecol. 102(2), 133–136 (1995).

23.

American College of Obstetricians and Gynecologists. ACOG committee opinion no. 426: the role of transvaginal ultrasonography in the evaluation of postmenopausal bleeding. Obstet. Gynecol. 113(2 Pt 1), 462–464 (2009).

24.

Epstein

. Management of postmenopausal bleeding in Sweden: a need for increased use of hydrosonography and hysteroscopy. Acta Obstet. Gynecol. Scand. 83(1), 89–95 (2004).

25.

Goldstein

Bree

Benson

Evaluation of the woman with postmenopausal bleeding: society of radiologists in ultrasound-sponsored consensus conference statement. J. Ultrasound Med. 18(1), 61–69 (2002).

26.

Ingui

Rogers

. Searching for clinical prediction rules in MEDLINE. J. Am. Med. Inform. Assoc. 8(4), 391–397 (2001).

27.

Hayden

Cote

Bombardier

. Evaluation of the quality of prognosis studies in systematic reviews. Ann. Intern. Med. 144(6), 427–437 (2006).

28.

Leushuis

van der Steeg

Steures

Prediction models in reproductive medicine: a critical appraisal. Hum. Reprod. Update 15(5), 537–552 (2009).

29.

Important systematic review of prediction models in reproductive medicine .

30.

McGinn

Guyatt

Wyer

Users' guides to the medical literature: XXII: how to use articles about clinical decision rules. Evidence-based medicine working group. JAMA 284(1), 79–84 (2000).

31.

Important overview on how to use clinical decision rules in clinical practice .

32.

Altman

Royston

. What do we mean by validating a prognostic model? Stat. Med. 19(4), 453–473 (2000).

33.

Reilly

Evans

. Translating clinical research into clinical practice: impact of using prediction rules to make decisions. Ann. Intern. Med. 144(3), 201–209 (2006).

34.

Steyerberg

. Calibration. In: Clinical Prediction Models — A Practical Approach to Development, Validation and Updating. Springer, Germany, 270–278 (2009).

35.

Steyerberg

. Discriminative ability. In: Clinical Prediction Models — A Practical Approach to Development, Validation and Updating. Springer, Germany, 260–269 (2009).

36.

Vergouwe

. Validation of clinical prediction models: theory and applications in testicular germ cell cancer (Dissertation). Rotterdam: Erasmus University, Rotterdam, The Netherlands (2003).

37.

Epstein

Skoog

Isberg

An algorithm including results of gray-scale and power Doppler ultrasound examination to predict endometrial malignancy in women with postmenopausal bleeding. Ultrasound Obstet. Gynecol. 20(4), 370–376 (2002).

38.

Randelzhofer

Prompeler

Sauerbrei

Value of sonomorphological criteria of the endometrium in women with postmenopausal bleeding: a multivariate analysis. Ultrasound Obstet. Gynecol. 19(1), 62–68 (2002).

39.

Bachmann

ter Riet

Clark

Probability analysis for diagnosis of endometrial hyperplasia and cancer in postmenopausal bleeding: an approach for a rational diagnostic workup. Acta Obstet. Gynecol. Scand. 82(6), 564–569 (2003).

40.

Bruchim

Biron-Shental

Altaras

Combination of endometrial thickness and time since menopause in predicting endometrial cancer in women with postmenopausal bleeding. J. Clin. Ultrasound 32(5), 219–224 (2004).

41.

Opmeer

van Doom

Heintz

Improving the existing diagnostic strategy by accounting for characteristics of the women in the diagnostic work up for postmenopausal bleeding. BJOG 114(1), 51–58 (2007).

42.

One of the two articles showing the best performance regarding discrimination and clinical usefulness in internal validation .

43.

Opolskiene

Sladkevicius

Valentin

. Ultrasound assessment of endometrial morphology and vascularity to predict endometrial malignancy in women with postmenopausal bleeding and sonographic endometrial thickness >or= 4.5 mm. Ultrasound Obstet. Gynecol. 30(3), 332–340 (2007).

44.

Burbos

Musonda

Giarenis

Predicting the risk of endometrial cancer in postmenopausal women presenting with vaginal bleeding: the Norwich DEFAB risk assessment tool. Br. J. Cancer 102(8), 1201–1206 (2010).

45.

Burbos

Musonda

Duncan

Estimating the risk of endometrial cancer in symptomatic postmenopausal women: a novel clinical prediction model based on patients' characteristics. Int. J. Gynecol. Cancer 21(3), 500–506 (2011).

46.

Opolskiene

Sladkevicius

Valentin

. Prediction of endometrial malignancy in women with postmenopausal bleeding and sonographic endometrial thickness ≥4.5 mm. Ultrasound Obstet. Gynecol. 37(2), 232–240 (2011).

47.

One of the two articles showing the best performance regarding discrimination and clinical usefulness in internal validation .

48.

Feldman

Cook

Harlow

Predicting endometrial cancer among older women who present with abnormal vaginal bleeding. Gynecol. Oncol. 56(3), 376–381 (1995).

49.

Weber

Belinson

Piedmonte

. Risk factors for endometrial hyperplasia and cancer among women with abnormal bleeding. Obstet. Gynecol. 93(4), 594–598 (1999).

50.

Khan

Bachmann

ter Riet

. Systematic reviews with individual patient data meta-analysis to evaluate diagnostic tests. Eur. J. Obstet. Gynecol. Reprod. Biol. 108(2), 121–125 (2003).

51.

Musonda

Burbos

Duncan

Comparing the performance of two clinical models in estimating the risk of endometrial cancer in symptomatic postmenopausal women. Eur. J. Obstet. Gynecol. Reprod. Biol (2011).

Prediction Models in Women with Postmenopausal Bleeding: A Systematic Review

Abstract

Keywords

Methods

Study identification

Study selection

Study quality assessment

Predictor variables

Model development assessment

Model performance

Results

Study identification & selection

Study characteristics

Study quality

Predictor variables

Phases of model development

Performance of the prediction models

Discussion

Future perspective

Footnotes

Appendix

References