Background: The aim of this study was to evaluate the role of maternal serum cancer antigen 125 (CA 125) and creatine kinase (CK) levels in predicting the outcome of pregnancies of unknown location (PUL).
Methods: Prospective observational study. Women classified as PUL were recruited. Final outcome of each PUL was established: failing PUL, intrauterine pregnancy (IUP), ectopic pregnancy (EP) or persisting PUL. The persisting PUL group almost certainly represent ultrasonically missed EP and were included in EP group. Serum CK and CA 125 were measured at 0 and 48 h. The values at presentation and the change in levels after 2 days were used in the analysis. We incorporated the most significant of variables into a multinomial logistic regression model to predict all outcomes. The performance of this model was evaluated using receiver operating characteristic (ROC) curves.
Results: In all, 4698 consecutive women were scanned; 403 were classified as PUL, 27 were lost to follow-up. Of the 376 women eligible, 297 had complete data and therefore were recruited. Mean age and mean gestation were 30.0 years and 43.3 days respectively. Final outcomes: 153 failing PUL (51.5%), 116 IUP (39.1%) and 28 EP (9.4%). Mean serum CK levels at 0 and 48 h were 88.5 and 86.8 IU/l respectively. Mean serum CA 125 levels at 0 and 48 h were 43.8 and 40.1 kIU/l respectively. 81.1% of women had CK and CA 125 ratios (CK 48 h/CK 0 h, CA 125 48 h/CA125 0 h) between 0.7 and 1.3. CA 125 ratio was the only significant variable in the three outcome groups (P < 0.0001). Logistic regression analysis incorporating CA 125 ratio gave an area under ROC curve of 0.782 (SE = 0.041) for failing PUL, 0.768 (SE = 0.043) for IUP and 0.560 (SE = 0.078) for EP. This model was capable of distinguishing failing PUL from IUP, but was not able to detect EP.
Conclusions: Absolute levels of serum CK and CA 125 at the defined times cannot be used to predict the outcome of PUL. Although the CA 125 ratio when incorporated into logistic regression model can distinguish failing PUL from IUP, its inability to detect the high risk PUL, i.e. the developing EP, renders it inappropriate for use in the clinical setting.