Serum progesterone levels could predict diagnosis, completion and complications of miscarriage
Tijdschriftbijdrage - Tijdschriftartikel
BACKGROUND: Low serum progesterone levels were strongly correlated with miscarriages in several publications and with completion of miscarriage in one paper. This study evaluated several parameters, predominantly serum progesterone, as predictors for miscarriages, their swift non-surgical completion and their complications.
BASIC PROCEDURES: Suspected or confirmed non-viable pregnancies with available concomitant serum progesterone measurements were retrospectively reviewed. The performance of serum progesterone, either alone or combined with other parameters, to predict viability, surgical removal and delay of non-surgical evacuation of non-viable pregnancy and complications, was analysed by logistic regression combined with Akaike and Bayesian information criteria, likelihood, receiver operated characteristic (ROC) curves, Mann-Whitney test and Fisher's exact test.
MAIN FINDINGS: From 151 included pregnancies, 104 (68.9%) were non-viable with 91 completions of miscarriage without surgery. The probability of viability was correlated linearly and curvilinearly with serum progesterone (p < 0.001). The probability of surgical removal, and the delay before non-surgical evacuation, showed a linear relationship with progesterone. No complication occurred when progesterone levels remained below 10 µg/L, while its rates were 9.5% of non-viable pregnancies with progesterone levels between 10 and 20 µg/L and 26.7% of cases with progesterone levels above 20 µg/L. Combined with progesterone, either "parity" or "history of miscarriage" improved the prediction of viability, "history of supra-isthmic uterine surgery" improved the prediction of surgery and "history of miscarriage" improved the prediction of delayed non-surgical evacuations.
CONCLUSION: Serum progesterone can probably predict the odds of miscarriages, surgical removal, delayed non-surgical evacuation and complications, with potential improvements when different predictors are combined.