OBJECTIVE To assess the performance of the human chorionic gonadotrophin (hCG) ratio (hCG 48 hours/hCG 0 hour) to predict spontaneous resolution of pregnancies of unknown location (PUL). METHODS Prospective cohort study. METHODS Teaching Hospital Early Pregnancy Unit. METHODS Women classified as having a PUL. METHODS The optimal cutoff value for hCG ratio (serum hCG at 48 hours/serum hCG at 0 hours) was calculated from data on 189 consecutive PULs (the 'training set'). This cutoff was tested prospectively on a further 200 consecutive PULs (the 'test set'). The hCG ratio was also compared to absolute levels of serum hCG at 0 and 48 hour for the prediction of failing PULs. METHODS hCG ratio in spontaneously resolving ('failing') PUL compared with those requiring intervention. Optimum cutoff determined and tested to predict spontaneously resolving PUL. Comparison of hCG ratio with absolute levels of serum hCG. RESULTS A total of 3996 consecutive women were scanned, of which 438 (11.0%) were classified as PULs. Complete data were available for 389 women: 189 in the training set and 200 in the test set. In the training set, there were 102 (54%) failing PUL, while 109 (55%) in the test set. hCG ratio of <0.87 predicted failing PUL, with a sensitivity of 93.1% (95% CI 85.9-97.0) and a specificity of 90.8% (95% CI 82.2-95.7) in the training set. In the test set, sensitivity was 92.7% (95% CI 85.6-96.5) and specificity was 96.7% (95% CI 90.0-99.1). The hCG ratio outperformed absolute serum hCG levels at 0 and 48 hours. CONCLUSIONS We have defined the optimal hCG ratio for the prediction of failing PUL. Using this cutoff, clinicians can safely adopt a noninterventional approach in women with PUL.