Nine studies were included. The total number of patients was unclear, as the studies contributed different numbers of patients to different tables depending on the review question to be addressed.
Accuracy of history and symptoms for pregnancy diagnosis.
A delayed menstrual period yielded statistically-significant results for predicting pregnancy in 3 out of 4 reported studies: positive LRs of 1.13 (95% confidence interval, CI: 1.05, 2.92), 1.56 (95% CI: 1.40, 1.74) and 2.06 (95% CI: 1.65, 2.57). However, the results were inconsistent and, therefore, represent an unreliable symptom of pregnancy. Typical early symptoms of pregnancy were more consistent across studies and slightly increased the likelihood of pregnancy: the positive LR was 2.43 (95% CI: 1.71, 3.44) for any symptoms and 2.70 (95% CI: 2.19, 3.33) for morning sickness. The absence of early symptoms, however, did not rule out pregnancy: the negative LR was 0.71 (95% CI: 0.67, 0.76) for absence of morning sickness and 0.63 (95% CI: 0.52, 0.77) for absence of any pregnancy symptoms. The use of birth control decreased the likelihood of pregnancy (pooled negative LR 0.29, 95% CI: 0.16, 0.53), but not sufficiently to rule it out. Patient suspicion of pregnancy statistically altered the likelihood of pregnancy, but not sufficiently to be reliable.
Accuracy of the physical examination.
No studies were found to have examined the inter- or intra-observer reliability. The most useful findings on physical diagnosis appeared to be the Chadwick sign (positive LR 28.7, 95% CI: 4.10, 200.00) and palpable uterine artery pulsation (positive LR 10.98, 95% CI: 5.63, 21.4), though the 2 studies from which these estimates were derived were of relatively lower methodological quality. In addition, the absence of these signs did not rule out pregnancy.
Accuracy of home pregnancy tests.
Narrative information was presented from 2 studies. In one study of 3 kits, positive and negative LRs of 2.5 and 0.29, respectively, were found. In the second study of 11 kits, the sensitivity and specificity were 100% under ideal laboratory conditions. However, a diagnostic study in 638 volunteers found 5 of the 11 kits had 100% specificity, whilst the others had specificities between 77 and 94%. Two kits had a high diagnostic sensitivity (greater than 90%), while 2 were found to have a low sensitivity (less than 10%).