Accuracy of single progesterone test to predict early pregnancy outcome in women with pain or bleeding: meta-analysis of cohort studies

Abstract

Objective: To determine the accuracy with which a single progesterone measurement in early pregnancy discriminates between viable and non-viable pregnancy.

Design: Systematic review and meta-analysis of diagnostic accuracy studies.

Data sources: Medline, Embase, CINAHL, Web of Science, ProQuest, Conference Proceedings Citation Index, and the Cochrane Library from inception until April 2012, plus reference lists of relevant studies.

Study selection: Studies were selected on the basis of participants (women with spontaneous pregnancy of less than 14 weeks of gestation); test (single serum progesterone measurement); outcome (viable intrauterine pregnancy, miscarriage, or ectopic pregnancy) diagnosed on the basis of combinations of pregnancy test, ultrasound scan, laparoscopy, and histological examination; design (cohort studies of test accuracy); and sufficient data being reported.

Results: 26 cohort studies, including 9436 pregnant women, were included, consisting of 7 studies in women with symptoms and inconclusive ultrasound assessment and 19 studies in women with symptoms alone. Among women with symptoms and inconclusive ultrasound assessments, the progesterone test (5 studies with 1998 participants and cut-off values from 3.2 to 6 ng/mL) predicted a non-viable pregnancy with pooled sensitivity of 74.6% (95% confidence interval 50.6% to 89.4%), specificity of 98.4% (90.9% to 99.7%), positive likelihood ratio of 45 (7.1 to 289), and negative likelihood ratio of 0.26 (0.12 to 0.57). The median prevalence of a non-viable pregnancy was 73.2%, and the probability of a non-viable pregnancy was raised to 99.2% if the progesterone was low. For women with symptoms alone, the progesterone test had a higher specificity when a threshold of 10 ng/mL was used (9 studies with 4689 participants) and predicted a non-viable pregnancy with pooled sensitivity of 66.5% (53.6% to 77.4%), specificity of 96.3% (91.1% to 98.5%), positive likelihood ratio of 18 (7.2 to 45), and negative likelihood ratio of 0.35 (0.24 to 0.50). The probability of a non-viable pregnancy was raised from 62.9% to 96.8%.

Conclusion: A single progesterone measurement for women in early pregnancy presenting with bleeding or pain and inconclusive ultrasound assessments can rule out a viable pregnancy.

Fig 1 Flow chart of study selection. *Three studies by Stovall et al reported same patients as study by McCord et al. Study by McCord was included, because it reported on more participants

Fig 2 Cumulative bar plot of methodological quality items across studies including women with pain or bleeding and inconclusive ultrasound (top) and women with pain or bleeding alone (bottom)

Fig 3 Forest plot of study results of progesterone test in women with pain or bleeding and inconclusive ultrasound assessment grouped according to outcome. FN=false negative; FP=false positive; TN=true negative; TP=true positive

Fig 4 Summary receiver operating characteristics plot of progesterone test at cut-off values between 3.2 and 6.4 ng/mL used to identify non-viable pregnancies in women with pain or bleeding and inconclusive ultrasound assessment (black dot=summary sensitivity and specificity; dotted region around it=95% confidence region)

Fig 5 Forest plot of study results of progesterone test at various cut-off values used to identify non-viable pregnancies in women with pain or bleeding alone. Study names with suffixes A to D reported accuracy of progesterone at more than one cut-off. FN=false negative; FP=false positive; TN=true negative; TP=true positive

Fig 6 Summary receiver operating characteristics plot of progesterone test at different cut-off values used to identify non-viable pregnancies in women with pain or bleeding alone