Background: The U.S. Preventive Services Task Force (USPSTF) does not currently have an active recommendation for preeclampsia screening. Preeclampsia is a complex disease occurring in the second half of pregnancy, and is estimated to affect nearly 4 percent of pregnancies in the United States. Nearly 9 percent of maternal deaths in the United States are directly attributed to preeclampsia and eclampsia, and it is a leading cause of induced preterm birth and low birth weight. Early detection through general or high-risk screening approaches may help reduce the health-related consequences, particularly for infants.
Purpose: We conducted a systematic review to assess the direct evidence of benefits and harms of preeclampsia screening on health outcomes; to evaluate the effectiveness of routine blood pressure and urine protein screening tests to identify women with preeclampsia; to estimate the accuracy of screening tests for proteinuria; and to evaluate the performance of multivariable risk assessment tools used during the first trimester to identify women at increased risk of preeclampsia as well as the potential harms of risk assessment.
Data Source: MEDLINE, PubMed, and Cochrane Central Register of Controlled Trials from 1990 through September 1, 2015. We included all references from the 1996 USPSTF recommendation and examined reference lists of relevant systematic reviews.
Study Selection: English-language trials and observational studies of screening effectiveness, test accuracy, and harms. Two investigators independently reviewed identified abstracts and full-text articles against a set of a priori inclusion and quality criteria.
Data Analysis: One investigator abstracted details about study design, patient population, setting, screening method, followup, and results. Two investigators independently applied prespecified criteria to rate study quality. Discrepancies were resolved through consensus, and poor-quality studies were excluded. Due to small numbers of studies and methodological shortcomings, meta-analysis was not attempted for any outcome measure other than urine protein:creatinine tests performed as point-of-care screening.
Results: A fair-quality randomized, controlled trial of 2,764 “low-risk” pregnant U.S. women found no statistically significant differences in health outcomes among women assigned to fewer prenatal screening visits compared with usual care at a large managed care organization in 1996 (mean number of visits, 12.0 vs. 14.7; p<0.001). A fair-quality before-after study of 1,952 low-income pregnant Hispanic women did not identify harms related to preeclampsia diagnosis and birth outcomes when protein urine screening was used for specific indications instead of on a routine basis in prenatal care. We found no evidence to evaluate the effectiveness of routine screening tests in identifying women with preeclampsia and limited evidence on various screening approaches for establishing the presence of proteinuria (a diagnostic criterion for preeclampsia). Fourteen diagnostic test accuracy studies (four good-quality, 10 fair-quality) compared point-of-care tests used to screen for proteinuria versus the gold standard (24-hour urine collection). Included studies of test accuracy were conducted in women with suspected preeclampsia, while studies with healthy, asymptomatic patients seeking routine care were lacking. Twelve studies evaluated the performance of protein:creatinine tests. High heterogeneity precluded pooling of test performance (k=11). Sensitivity for the protein:creatinine test ranged from 0.65 to 0.96 (I2=80.5%; 11 studies) and specificity ranged from 0.49 to 1.00 (I2=91.8%; 11 studies). Statistical heterogeneity of test sensitivity was partly explained by differences in the study populations; studies with a positive protein dipstick result as an inclusion criterion had higher sensitivity (p<0.05). Two studies of the albumin:creatinine spot test had high sensitivity (≥0.94, [95% confidence interval, 0.75 to 1.00]). Four studies of quantitatively read protein dipstick tests had widely variable sensitivity (0.22 to 1.00) and specificity (0.36 to 1.00). Four studies validated five first-trimester risk assessment models with good-to-excellent discrimination, primarily for predicting early-onset preeclampsia requiring delivery. No externally validated multivariable risk prediction models were based only on patient history measures that could be collected in a routine prenatal care visit; all included serum markers and uterine artery Doppler ultrasound measure of the pulsatility index, or both. Five models had good discrimination of preeclampsia cases (c-statistic, >0.80) but very low positive predictive values and did not provide necessary information on model calibration.
Conclusions: Changes in diagnostic criteria, patient demographics, and treatment recommendations affect the applicability of previous trials, precluding conclusions about the optimal screening approach. Most studies for detecting proteinuria, one of the diagnostic criteria for preeclampsia, tested the protein:creatinine ratio in urine samples; however, all studies were among patients with prescreened suspicion of preeclampsia and none evaluated the performance of repeat testing of urine protein for screening. Due to limited and variable evidence, different urine protein screening tests cannot be compared. There was no clear evidence of the performance, clinical benefits, or harms of any externally validated models for risk prediction, and the clinical performance and impact of risk prediction models could not be extrapolated to relevant patient settings. Current screening practices are considered routine and represent relatively minor burdens to patients, clinicians, and health care systems, but evidence is limited for determining the benefits and harms of preeclampsia screening.