Twenty-three studies were included: 16 cohort studies and 7 case- control studies. Twenty-two studies reported on the performance of serum CA-125 measurement in the diagnosis of any grade of endometriosis, while 18 studies reported on its performance in the diagnosis of severe endometriosis (grade III/IV).
Diagnosis of any form of endometriosis (n=22).
The sensitivity ranged from 0% (specificity 100%) to 100% (specificity 93%) and the specificity from 38% (sensitivity 50%) to 100% (sensitivity 0 to 85%). Significant heterogeneity was present for both sensitivity and specificity (P<0.001). The logistic regression analysis showed that the performance of serum CA-125 measurement differed significantly between cohort and case- control studies (ratio of the diagnostic odds ratios = 9.6, P=0.001), showing that the diagnostic odds ratio was higher in case-control studies than in cohort studies. Further analysis was limited to cohort studies, as these are considered superior to case-control studies. Heterogeneity in the sensitivity and specificity remained; Spearman's correlation coefficient was - 0.76. A ROC curve was estimated and produced; this showed a low diagnostic performance. When the sensitivity was increased to 50% the specificity decreased to 72%.
Diagnosis of severe endometriosis (n=18).
The sensitivity ranged from 0% (specificity 80%) to 100% (specificity 76%) and the specificity from 44% (sensitivity 60%) to 95% (sensitivity 53%). Significant heterogeneity was present (P<0.001), therefore the summary sensitivity and specificity were not calculated. The logistic regression analysis showed no difference in the performance of serum CA-125 measurement as reported by studies with different designs; Spearman's correlation coefficient was -0.59. A summary ROC curve was estimated; this showed that the capacity of serum CA-125 measurement in the diagnosis of severe endometriosis was better than for any type of endometriosis. For a specificity of 89% the sensitivity was 47%. When the sensitivity was increased to 60% the specificity decreased to 81%.