Objective
We sought to evaluate whether hysteroscopy in patients with endometrial cancer had an effect on disease stage or mortality.
Study Design
This was a retrospective cohort analysis of data linked between a registry of women diagnosed with endometrial cancer and physician billing data on hysteroscopy.
Results
A 99.8% match rate was obtained. Eighty-five percent of cases had complete data on staging. Of these 1972 cases, 672 (34.1%) had undergone hysteroscopy. There was no difference in stage III disease between the hysteroscopy (7.1%) vs no hysteroscopy (6.5%) group ( P = .38). There was also no difference in death rates, 13.2% vs 15.2% ( P = .25), or in the proportion of women dying of female genital organ cancer, 46.1% vs 42.1% ( P = .53), respectively.
Conclusion
Hysteroscopy is not associated with a higher rate of stage III disease or mortality. It allows for accurate diagnosis with direct visualization and biopsy, and should be considered a safe diagnostic tool.
Hysteroscopy is widely used for the diagnosis of endometrial hyperplasia or carcinoma. The sensitivity has been estimated to be 86.4%. In comparison, Pipelle endometrial biopsies have been reported to have a sensitivity of 67-83.5%. It has also been shown that hysteroscopically guided curettage has higher accuracy than dilatation and curettage (D&C) alone. However, concern exists that the introduction of high-pressure gas or fluid into the uterine cavity, to produce distension, could facilitate the dissemination of malignant cells in patients with endometrial cancer. It has been found that the distention medium can begin draining transcervically and transtubally when pressures reach approximately 100-150 mm Hg. Thus, in the patient with endometrial cancer, there is the theoretical possibility that malignant cells might be dispersed into the fallopian tubes and the abdominal cavity following hysteroscopy. The evidence associating peritoneal dissemination with hysteroscopy and ultimately its possible effect on endometrial cancer mortality outcomes is inconclusive.
This study was undertaken to evaluate whether preoperative hysteroscopy performed in patients with endometrial cancer had an effect on subsequent surgical staging, and ultimately mortality. It is a retrospective cohort analysis of data on women who had been diagnosed with endometrial cancer over a period of 10 years (1997 through 2006).
Materials and Methods
Sources of data for this study were the Alberta Cancer Registry and physician billing data from Alberta Health and Wellness. The Alberta Cancer Registry captures data on all patients diagnosed with a cancer that is identified by the International Classification for Diseases for Oncology . Data on women with endometrial cancer were retrieved from the registry from Jan. 1, 1997, through Dec. 31, 2006. This start time was chosen principally because the Alberta Cancer Registry had undergone a number of coding changes prior to this time period. It was not until 1997 that endometrioid adenocarcinoma was differentiated from other more aggressive nonadenocarcinoma histologic types, such as, clear cell and papillary serous. A second benefit to this time period was the introduction of new staging criteria for endometrial cancer that took place in 1989. Staging after that change included surgical observations. Besides diagnosis, the Alberta Cancer Registry provided further information on pathology (histologic cell type and staging) and vital statistics. Loss to follow-up within the Alberta Cancer Registry is minimal, as each month linkage is made to Alberta Vital Statistics to identify those cancer patients who have died and the cause of death. In addition, there is a link made annually with the Canadian National Mortality database. However, patients who die outside Canada could be potentially missed.
Cases were selected from the Alberta Cancer Registry based on the diagnosis of endometrial cancer as defined by a positive biopsy. Staging was reported by the treating physician. In Alberta, pelvic washings are considered the standard of care. The majority of cases would have also included pelvic lymph node sampling. For the few cases missing node sampling, the performance of pelvic washings alone would have dictated the classification of stage IIIa disease based on the International Federation of Gynecology and Obstetrics (FIGO) 1989 staging criteria.
A thorough review of this dataset was then performed to eliminate transcription errors and to clarify staging. When staging information within the dataset was missing or not interpretable, cross-reference to other variables (ie, text entered by the physician) in the wider electronic database allowed for clarification in many cases. Finally, any staging based on the American Joint Committee’s Cancer TNM staging manual (fifth and sixth editions) was translated to the FIGO 1989 staging criteria.
Alberta Health and Wellness administers a universal health care plan that covers all Albertans, and through this, physicians are reimbursed for all procedures performed in the province. The Alberta Health and Wellness dataset provided information on diagnostic hysteroscopy (specifically by fee code 80.81) and patient characteristics (age). This code would have captured hysteroscopy performed in both the office setting and operating room. When hysteroscopy was not performed, diagnoses would have been made by either office endometrial biopsy or D&C.
The cases from the cancer registry were then linked to the Alberta Health and Wellness dataset. Information on whether a diagnostic hysteroscopy was performed for each patient was abstracted. Data linkage went back as far as January 1996, to include information on cases that were pulled from the Alberta Cancer Registry for the year of 1997. This would represent hysteroscopy that had been performed within 1 year, isolating that which was used for the diagnosis of endometrial cancer. Data linkage was performed using at least 2 patient-specific variables within each dataset, and all identifying information was removed from the linked dataset before release for analysis.
Women with stage IV disease were excluded, since diagnosis with an endometrial sample may not have been required, as disease may have been more clinically apparent at presentation. As the survival in this group is also close to 0, this would have affected our analysis by disproportionably decreasing the survival for women who had not had hysteroscopy. Cases were also excluded if histology indicated a nonadenocarcinoma type of cancer, such as squamous, clear cell, papillary serous, and/or undifferentiated carcinoma. This eliminated advanced disease that was likely due to the inherent disease process of a more aggressive tumor type, rather than the diagnostic technique used.
There were 2 primary outcomes in this study: staging and overall survival. Staging ≥IIIa would indicate that the carcinoma had extended out of the uterus. If peritoneal dissemination via the fallopian tubes occurs during hysteroscopy, staging should indicate at least stage III. Therefore, the study hypothesis was that hysteroscopy would be associated with an increased frequency of stage III disease and endometrial cancer–related deaths.
Review of the literature suggested that 10% of those women with endometrial cancer who undergo hysteroscopy for diagnostic purposes and 5% of those who do not undergo hysteroscopy will have stage III disease. From this, it was calculated that a sample size of 387 and 581 women who have and have not undergone hysteroscopy, respectively, would be needed to detect a difference in staging with a power of 80% and an alpha error of 0.05. These numbers were easily attained from the provincial registries.
Data were summarized using proportions for categorical variables, and both mean and median for continuous variables. Bivariate analyses were performed to compare the hysteroscopy groups in regard to staging and other possible prognostic risk factors. Comparisons were made using Pearson χ 2 test for categorical variables. Mantel-Haenszel χ 2 test was used to identify trends in proportions. A Student t test was used to compare continuous variables if they were normally distributed; otherwise, the Wilcoxon rank sum test was used. P ≤ .05 was considered to be statistically significant. All statistical calculations were performed using the SAS program package (SAS Institute, Cary, NC).
Data were managed according to the Alberta Privacy Legislation. Data linkage between Alberta’s Cancer Registry and the Health and Wellness dataset was performed by Alberta Health and Wellness. This process did not compromise patient confidentiality as only anonymous data were received by the research team. Ethical approval was obtained from the Conjoint Health Research Ethics Board at the University of Calgary (Ethics Review Board no: E-20934).
Results
The initial data request from the Alberta Cancer Board had provided 2331 cases that met both inclusion and exclusion criteria. A 99.8% match rate was obtained when linkage was performed with the Alberta Health and Wellness claims data. Approximately 85% of these cases had complete and appropriate data on staging (350 were missing staging completely, while 5 had staging complicated by coexistent cancer, a different type of cancer, or a benign condition). Of these 1972 cases with complete staging, 672 (34.1%) had undergone hysteroscopy.
There was an increase in the utilization of hysteroscopy over the study period, and in the number of cases diagnosed. In 1997, 40 hysteroscopies were performed in 154 women who were diagnosed with endometrial cancer (26.0%), whereas, in 2006, 95 hysteroscopies were performed in 232 women (40.9%).
The mean age was 62.0 and 60.7 years, respectively, for women who had undergone hysteroscopy and those who had not (Wilcoxon rank sum, P = .01).
The frequencies of FIGO stages in the 2 groups are shown in Table 1 . The rates of stage III disease did not differ statistically between those who had and did not have hysteroscopy (relative risk, 1.16; 95% confidence interval [CI], 0.83–1.62; P = .38).
| Hysteroscopy | |||
|---|---|---|---|
| Stage | Yes n = 672 No. (%) | No n = 1300 No. (%) | Statistical test |
| Stage <III | 621 (92.9) | 1215 (93.5) | χ 2 P = .38 |
| Stage III | 51 (7.1) | 85 (6.5) | RR, 1.16; 95% CI, 0.83–1.62 |
| Details of FIGO stage | |||
| FIGO 0 | 1 (0.2) | 3 (0.2) | |
| FIGO I a | 67 (10.0) | 112 (8.6) | |
| FIGO Ia | 133 (19.8) | 265 (20.4) | |
| FIGO Ib | 182 (27.1) | 447 (34.4) | |
| FIGO Ic | 116 (17.3) | 201 (15.5) | |
| FIGO II a | 2 (0.3) | 19 (1.5) | |
| FIGO IIa | 49 (7.3) | 55 (4.2) | |
| FIGO IIb | 71 (10.6) | 113 (8.7) | |
| FIGO III a | 1 (0.2) | 3 (0.2) | |
| FIGO IIIa | 21 (3.1) | 54 (4.2) | |
| FIGO IIIb | 3 (0.5) | 5 (0.4) | |
| FIGO IIIc | 26 (3.9) | 23 (1.8) | |
There was also no statistically significant difference found in death rates for those who had a hysteroscopy (13.2%) compared to those who had not (15.2%) (relative risk, 0.87; 95% CI, 0.69–1.10; P = .25) ( Table 2 ). The mean years of follow-up for each group were 4.0 (0.03-9.97) and 4.4 (0.02-10.67), respectively. Survival analysis was not possible because of the relatively fewer numbers of deaths, resulting in high censor rate in the analysis. When evaluating only those women who had died, statistically significant differences were found neither in the proportion of stage III disease ( P = .10), nor in the proportion of women dying of female genital organ cancer between hysteroscopy groups ( P = .53) ( Table 3 ).
