This review concluded that accuracy of CA 125 for the diagnosis of ovarian tumours was high. The review suffered from a number of limitations, including a restricted literature search and considerable heterogeneity between studies that was not appropriately investigated. The pooled results did not support the conclusion that the accuracy of CA 125 was high.

To estimate the accuracy of the cancer antigen (CA) 125 assay for the diagnosis of ovarian tumours.

MEDLINE, CANCERLIT, LILACS, EMBASE and The Cochrane Library were searched for studies published from 1985 to December 2007. Search terms were reported and included a diagnostic filter. Reference lists of primary studies were screened to identify additional studies. No language restrictions were applied.

Diagnostic accuracy studies that evaluated CA 125 at a threshold of 35U/ml for the diagnosis of ovarian tumours in women with clinically suspected adnexal masses treated surgically for ovarian tumours were eligible for inclusion. Studies were required to use histological analysis of standard paraffin-embedded sections as the reference standard and to classify specimens as benign, borderline or malignant. Studies had to report sufficient data to construct a 2x2 table of test performance. Mean age, where reported, ranged from 33 to 48 years.

Three reviewers independently assessed studies for inclusion. Disagreements were resolved through consensus or by referral to a fourth reviewer.

Three reviewers independently assessed study quality according to the following criteria: method of data collection and patient selection; description of CA 125 levels; description of the histological reference standard; and the presence of verification bias.

Three reviewers independently extracted data as 2x2 tables of test performance with malignant or borderline disease considered as positive and benign lesions considered as negative. These data were used to calculate sensitivity and specificity. When 2x2 tables contained a 0 cell, 0.5 was added to enable calculations. Studies that contained two 0 cells were excluded. Disagreements were resolved through consensus.

Pooled sensitivity and specificity together with 95% confidence intervals (CIs) were calculated using random-effects models. Where sensitivity and specificity were thought to be correlated, a summary receiver operating characteristic (SROC) analysis using the Moses-Littenberg model was undertaken. Heterogeneity in sensitivity and specificity was assessed using the Cochrane Q test and I² statistic. Sensitivity analysis was undertaken by excluding studies that fulfilled less than 50% of the quality criteria. Publication bias was assessed using funnel plots based on the diagnostic odds ratio (DOR) and sample size.

Seventeen studies (n=2,374) were included. Only three studies were considered to be of high methodological quality as they fulfilled at least 55% of the quality criteria. All studies included small numbers of patients. No studies were blinded. All studies failed to provide sufficient details on the included patients. Only three studies were reported to be prospective and consecutive, one was not and all other studies did not provide information on this. All studies avoided verification bias and provided sufficient information on the index test and reference standard.

Sensitivity ranged from 28% to 100%. Pooled sensitivity was 80% (95% CI 76% to 82%). There was strong evidence of heterogeneity (p=0.001, I²=89%).

Specificity ranged from 46% to 98%. Pooled specificity was 75% (95% CI 73% to 77%). There was strong evidence of heterogeneity (p=0.001, I²=96%).

Restriction of the analysis to the three studies judged to be of high methodological quality showed similar results.

The accuracy of CA 125 for diagnosis of ovarian tumours was high and very important in helping the surgeon to decide what type of surgery should be performed.

The review addressed a focused question supported by clearly defined inclusion criteria. The literature search included a broad range of databases, but the use of a diagnostic filter meant that relevant studies were likely to have been missed. No language restrictions were applied, which minimised the risk of language bias. The restriction to published studies meant that publication bias may have been a problem. Publication bias was assessed in the review, but the methods used were not appropriate for diagnostic accuracy studies. Appropriate steps were taken to minimise bias and errors at all stages of the review process. Study quality was assessed using appropriate criteria, but the results were not appropriately considered in the analysis. There was considerable heterogeneity between studies, so it was questionable whether pooling data was appropriate (especially so given that reasons for the observed heterogeneity were not explored). The pooled results did not support the authors' conclusions that the accuracy of CA 125 for the diagnosis of ovarian tumours was high. This, combined with the limitations discussed above, make the authors conclusions unlikely to be reliable.

Practice: The authors stated that the CA125 antigen at levels of at least 35U/ml was a useful preoperative test for predicting a benign or malignant pelvic mass.

Research: The authors stated that blinded prospective studies were needed to investigate the accuracy of CA 125 in association with transvaginal ultrasound in the diagnosis of ovarian tumours.

Not stated.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.