Caution necessary when interpreting results of outpatient endometrial sampling - Patient-Oriented Evidence that Matters

Journal of Family Practice, July, 2002 by M. Elizabeth Swenor, Ellen G. Smith

Clark TJ, Mann CH, Shah N, Khan KS, Song F, Gupta JK. Accuracy of outpatient endometrial biopsy in the diagnosis of endometrial cancer: a systematic quantitative review. Br J Obstet Gynaecol 2002; 109:313-21.

* BACKGROUND Outpatient endometrial sampling in symptomatic women with abnormal vaginal bleeding is a common practice in primary care. Results from existing studies evaluating various out patient office-based endometrial sampling techniques are inconsistent.

* POPULATION STUDIED The goal of this systematic quantitative review of the published literature was to determine the accuracy of outpatient endometrial biopsy in detecting endometrial cancer. The authors searched general bibliographic databases (MEDLINE and EMBASE) without language restrictions from 1980 through 1999 for articles comparing outpatient endometrial biopsy results with a reference (gold) standard (most commonly dilation and curettage, hysterectomy, or guided biopsy). Of 1369 trials initially screened, only 11 that were either prospective observational or comparative cross-sectional studies met the inclusion criteria. These 11 trials enrolled a total of 1013 pre- and postmenopausal women with abnormal uterine bleeding; post-menopausal women represented nearly 80% of the study subjects. No additional patient information was reported. The prevalence of endometrial cancer in the study population was 6.3%.

* STUDY DESIGN AND VALIDITY The small number and poor quality of the existing studies significantly limited this analysis. Two authors independently reviewed the studies for inclusion, and disagreement was resolved by consensus or arbitration by a third reviewer. Prospective and consecutive recruitment of eligible women were considered adequate for inclusion, whereas convenience sampling was considered inadequate. Blinding was considered adequate if the pathologists providing gold standard histological diagnoses were unaware of the results of the outpatient biopsy and inadequate if they were aware of the results. A decision to perform a reference test only in response to the results of an outpatient biopsy was considered inadequate.

Seven of the 11 selected studies enrolled nonconsecutive patients and only 2 of the 11 studies reported that outcomes were assessed blindly; blinding was not reported by the other studies. The nonblinding of pathologists interpreting the reference standards may have overinflated the sensitivities reported in the individual trials and thus biased the overall positive likelihood ratio(s) of

the individual and combined sampling techniques.

In this review, only results from studies evaluating the Pipelle outpatient device have adequate numbers of patients included to make a valid assessment, although the authors provided combined data on all the devices used as well. A funnel plot evaluation of the included studies found little evidence for publication bias.

* OUTCOMES MEASURED The primary outcome measure was the accuracy with which endometrial cancer was diagnosed by the various sampling techniques. Secondary outcomes were device failures and rates of inadequate specimens.

* RESULTS The pooled likelihood ratios for endometrial cancer using the Pipelle outpatient device with adequate endometrial sampling were 64.6 (95% confidence interval [CI], 22.3-187.1) for positive results and 0.1 (95% CI, 0.04-0.28) for negative results. The posttest probability given the initial prevalence of 6.3% with a positive outpatient test was 81.3% (95% CI, 52.4-94.4) and decreased to 0.7% (95% CI, 0.2-2.4) for a negative test. Inadequate samples were considered as negative results, which increased the accuracy. The overall failure rate (inability to perform the procedure for one reason or another) for outpatient biopsy was 7%.

Three endometrial cancers were missed with adequate endometrial samples and 1 cancer with an inadequate sample (false-negative rate = 0.4%). The pretest probability (prevalence) of endometrial cancer for all women in the study was 6.3%, which is lower than the commonly reported prevalence of 15% in postmenopausal women with abnormal vaginal bleeding. An increased pretest probability increases the false-negative rate, making a negative result less reliable.

RECOMMENDATIONS FOR CLINICAL PRACTICE

Caution is necessary when using office-based endometrial sampling techniques, including the Pipelle, to evaluate women with abnormal uterine bleeding. An abnormal histological finding is highly accurate and likely to represent true disease. Negative results, including inadequate sampling, must be interpreted with caution, because the false-negative rate for excluding endometrial cancer reported in this analysis was 4/1000 women sampled. Many clinicians and their patients may find this false-negative rate clinically unacceptable, while others may find reassurance from a "low-risk" assessment. In cases of abnormal uterine bleeding in which symptoms persist despite a negative biopsy, further evaluation and input from individual patients is recommended.