Pattern Recognition to Prognosticate Endometrial Cancer: The Science Behind the Art of Office Hysteroscopy—A Retrospective Study

A Predictive Model for Endometrial Carcinoma Based on Hysteroscopic Data

Objective

The purpose is to establish a model to predict endometrial carcinoma and assess its value in the preliminary diagnosis of endometrial carcinoma.

Methods

The data of 381 patients undergoing hysteroscopy were incorporated into the model, including 282 cases in the training cohort and 99 cases in the validation cohort. Significant morphological indexes were selected using the chi-square test and subjected to the binary logistic regression analysis. Besides, the scoring interval was set, and the nomogram of the prediction model was established. Model calibration curves were drawn using the data from the validation cohort. The study was approved by the Ethics Committee of the Affiliated Sir Run Run Hospital of Nanjing Medical University, and written informed consent was obtained from the patients.

Results

The sensitivity, specificity, positive predictive value, and negative predictive value of the model were 96.7%, 92.3%, 77.3%, and 99.0%, respectively. Analysis of the receiver operating characteristic curve in the training cohort showed an area under the curve of 0.984 (95% CI: 0.974-0.995). The receiver operating characteristic curve in the validation cohort revealed an area under the curve of 0.976 (95% CI: 0.950-1.000). The calibration curve indicated that the probability in the actual setting was consistent with that predicted by the nomogram in the training cohort.

Conclusion

Our model has high sensitivity and specificity in predicting endometrial carcinoma, and helps clinicians to make accurate diagnosis.

The use of hysteroscopy in endometrial cancer: old questions and novel challenges

Endometrial cancer is the most common gynecological malignancy with a relatively good overall prognosis. It traditionally has two subtypes: type 1 (endometrioid carcinoma) and type 2 (non-endometrioid carcinoma). The prognosis is excellent for stage I endometrioid cancer, with a 5-year survival rate of 96%. However, the prognosis is much worse for women with high-risk endometrial cancer. Effective preoperative staging is important in order to tailor treatment and achieve optimal long-term survival. The majority of asymptomatic polyps detected by ultrasound are treated surgically. Conventionally, dilatation and curettage was performed to obtain a histological diagnosis, but nowadays hysteroscopy with biopsy is starting to be considered as the gold standard. Hysteroscopic resection seems to reduce the risk of underdiagnosed (atypical endometrial hyperplasia) endometrial cancer. To avoid the spread of malignant cells, hysteroscopy should be performed with concern to keep intrauterine pressure low. In comparison with cervical injection, the hysteroscopic method has a better detection rate in the para-aortic area during sentinel lymph node mapping. In the assessment of cervical involvement, the accuracy of magnetic resonance imaging is significantly higher than the accuracy of hysteroscopy. In fertility-sparing cases, hysteroscopic endometrium resection with progesterone therapy is an acceptable option.

Preoperative assessment of endometrial cancer

Endometrial cancer (EC), the most common among gynaecological malignancies occurs predominantly after the menopause. The diagnosis is most commonly (in about 75-80%) set up at early stage when surgical therapy and if necessary postoperative radiotherapy results in an excellent prognosis, with a 90-95% 5-year overall survival (OS) and a locoregional recurrence rates of 4-8%. Accurate preoperative assessment of the lymph nodes would ideally identify those patients with advanced stage disease, who might benefit from more extensive surgical procedures and adjuvant therapies. magnetic resonance imaging (MRI), hysteroscopic excisional biopsy (HEB) and high resolution 2D or 3D ultrasound performed by expert operator are considered to add valuable information for preoperative staging of EC. The use of biomarkers could be beneficial in decreasing inter-observer variability between the histology of the diagnostic specimen and the final operative sample, as well as to avoid overtreatment in a part of the high-grade tumors with excellent prognosis. The goal of surgical management of EC is to remove the primary tumor and to identify definite prognostic factors to determine whether adjuvant therapy is required. Extended surgery, including para-aortic lymphadenectomy has a significant morbidity, and with the future selection of cases it can be safely avoided, we can decrease complications without compromising oncological safety. In the future, the importance of hysteroscopy guided sampling may increase to gain a representative sample for biomarker detection.

Establishing patterns on hysteroscopy in abnormal uterine bleeding (AUB)

Introduction:

Pattern recognition of various phases of normal endometrium and endometrial pathologies during hysteroscopy has many advantages. It would help to triage women with AUB, so as to be selective with biopsies and curettages. Recognition of normal variant or benign lesion would reduce burden to the pathologist by decreasing the number of unnecessary sampling. It will also decreases anxiety of the patient as the report/prognostication can be instant in many cases.

Material and methods:

This prospective, double blind, correlation study was carried out in the teaching hospital with a sample population of 70 women presenting with AUB who underwent hysteroscopy and endometrial sampling. We identified patterns of endometrium which can used to predict six endometrial pathologies which were later correlated with the final histological diagnosis.

Results:

There was good correlation between hysteroscopic patterns and histopathology report, 33% of starry sky appearance correlated with atrophic endometrium, 87% of tongue shaped projections correlated with endometrial polyp, 44.4% of pebble stone appearance correlated with myomatous polyp, 50% of polypoidal pattern correlated with endometrial hyperplasia. 100% correlation was seen in strawberry appearance, pattern for secretory endometrium and cerebroid appearance which was pattern designated to endometrial carcinoma.

Conclusion:

Hysteroscopic pattern recognition is a useful concept to triage women who require sampling for histopathological diagnosis.