Second-trimester urine nephrin:creatinine ratio versus soluble fms-like tyrosine kinase-1:placental growth factor ratio for prediction of preeclampsia among asymptomatic women

This prospective observational study compare urine nephrin:creatinine ratio (NCR, ng/mg) with serum soluble fms-like tyrosine kinase-1:placental growth factor ratio (FPR, pg/pg) for preeclampsia (PE) prediction among unselected asymptomatic pregnant women in 2nd trimester. NCR and FPR were determined in 254 paired urine/blood samples collected simultaneously from 254 women at median gestational week (GW) 24 (range, 22–27) without hypertension or significant proteinuria in pregnancy (SPIP). Fifteen (5.9%) developed SPIP and hypertension at GW 34.0 (26.0–38.6) and 35.3 (27.6–38.6), respectively, and were diagnosed with PE at GW 35.7 (27.6–38.6). The 90th percentile level determined in 239 women normotensive throughout pregnancy gave NCR (139) sensitivity and positive predictive values (PPV) of 60% (9/15) and 27% (9/33), while those for serum FPR (4.85) were 40% (6/15) and 20% (6/30), respectively. Relative risks (95%CI) of later PE were 10.0 (3.82–26.4; 27% [9/33] vs. 2.7% [6/221]) and 4.98 (1.91–13.0; 20% [6/30] vs. 4.0% [9/224]) for NCR-positive and FPR-positive women, respectively. Cut-offs suggested by ROC gave NCR (86.6) sensitivity and PPV of 87% (13/15) and 17% (13/79), and FPR (8.8) values of 40% (6/15) and 40% (6/15), respectively. Thus, 2nd trimester NCR was superior to FPR for PE prediction.

The median values were 1414 vs. 813 pg/mL for sFlit-1 levels and 336 vs. 497 pg/mL for PlGF levels determined at around GW 24 in 15 vs. 239 women that did and did not later develop PE, respectively, in this study. These levels were similar to those determined in women with similar clinical conditions in a previous study by Levine et al. 16 , which used the identical assay kit to the present study. Therefore, absolute FPR levels around GW 25 in the two clinical conditions in this study were also similar to those reported by Levine et al. 15 . However, our absolute sFlit-1 value was somewhat lower and the PlGF value was similar to those proposed as normal values by Verlohhren et al. 26 determined by different assay systems 26 ; median sFlit-1 values were 1299 and 1355 pg/mL at GW 20-23 and GW 24-28, respectively, and corresponding PlGF values were 264 and 465 pg/mL, respectively, in the report by Verlohhren et al. 26 . Therefore, the assay systems used by Verlohhren et al. 26 indicated higher normal FPR levels than our results: the median FPR levels were 4.92 and 3.06 at GW 20-23 and GW 24-28, respectively 26 , compared to 1.65 at GW 25 (22)(23)(24)(25)(26)(27) in the present study.
The 2 nd trimester NCR around GW 24 appeared to be superior to simultaneous FPR with regard to the detection rate of women with later development of PE in our setting. Women with NCR > 90 th percentile value had RR of PE = 10.0 (3.82-26.4) compared to women with NCR < 90 th percentile value; the corresponding RR was     Table 2 for their screening characteristics. 4.98 (1.91-13.0) for FPR in this study. In addition, AUC of ROC was greater for NCR than for FPR, and the NCR cut-off suggested by the ROC yielded high sensitivity of 87%. In our previous study using longitudinal urine samples from women that did and did not develop PE later 6 , the NCR did not change significantly in pregnant women that remained normotensive, while it increased with advancing gestation prior to the later development of PE, suggesting the possibility of using NCR as an urine biomarker of PE 6 . Therefore, NCR was considered worth comparing with the well-known biomarker, FPR, and the results indicated that NCR could be a clinically useful biomarker and may not be inferior to FPR. Screening characteristics of urinary nephrin for PE were examined in two previous studies with promising results 3,5 ; sensitivity and specificity were 73% and 79%, respectively, in one study 3 , and 57% and 58%, respectively, in the other study for prediction of PE 5 . However, these findings should be confirmed in larger studies before clinical use, as performed for FPR 24,26 . Unexpectedly, the 2 nd trimester NCR levels were independent of the 2 nd trimester sFlit-1, PlGR, and FPR (Fig. 5A). As exogenously administered sFlit-1 induces not only hypertension but also proteinuria associated with glomerular endotheliosis in pregnant rats 14 and as increased NCR was associated with increased proteinuria (Fig. 4A), a significant correlation was expected between NCR and sFlit-1/FPR level. However, there were no significant correlations between 2 nd trimester NCR and 2 nd trimester sFlit-1/FPR level in this study. These unexpected results gave the combined use of NCR and FPR a high PPV of 80% for prediction of PE in this study ( Fig. 5C). This was reasonable based on the findings that NCR was more predictive for SPIP than hypertension onset and that FPR was more predictive for hypertension than SPIP onset (Fig. 4). Thus, NCR and FPR had complementary roles to each other in detection of women with increased risk of PE.

Study limitations
As our institution is a tertiary centre managing mainly women at higher risk, the frequency of PE (5.9%, 15/254) in this study was somewhat higher than the prevalence rate of 2.3% in the general Japanese population 27 . The number of women aged ≥ 35 years (44% [112/254)), with twin pregnancies (13% [32/254]), and with complications including DM/GDM and connective tissue diseases (approximately 20%) were greater in this study than in the general population. As all of these are risk factors for PE [27][28][29] , this observation explained the somewhat higher frequency of PE in this study, and therefore our study population did not represent the general population.
In conclusion, non-invasive urine test of NCR around GW 25 was compared with simultaneous serum FPR with regard to predictability of later development of PE. Results were promising and larger prospective studies are warranted regarding the ability of NCR for prediction of PE.

Methods
This study was conducted in accordance with the principles of the Declaration of Helsinki and with the approval of the Institutional Review Board of Hokkaido University Hospital (013-3999, April 30, 2014), a tertiary teaching hospital managing mainly high-risk pregnant women.
Participants. During a study period between from May 2014 to June 2016, 827 women received antenatal care and gave birth at Hokkaido University Hospital. Among them, 540 (65%) women gave written informed consent to participate in this study and provided at least one spot urine specimen with or without simultaneously collected blood specimen during pregnancy. From these 540 women, 97 women were first excluded from the present analyses: 15 women with known hypertension (pre-existing hypertension) at the establishment of the current pregnancy and 82 women that provided urine samples only collected during the 1 st or 3 rd trimester. Then, from the 443 women with availability of 2 nd trimester urine specimens but without chronic hypertension, the following 189 women were excluded: one woman that exhibited hypertension before reaching GW 20, three women that exhibited new onset of hypertension and or SPIP at the time of urine collection, and 185 women whose blood specimen collected simultaneously with urine collection was not available. Thus, the remaining 254 women fulfilling the following two criteria were finally enrolled in this study: (1) no known pre-existing hypertension, and (2) no hypertension or SPIP at the simultaneous urine and blood sampling during the 2 nd trimester. Thirty-three of the 254 women (13%) were also enrolled in our previous study 6 .
SPIP was defined as PCR (mg/mg) > 0.27 (corresponding to 30 mg/mmol) in spot urine specimens. Hypertension was diagnosed in women with systolic blood pressure ≥ 140 mmHg and/or diastolic blood pressure ≥ 90 mmHg on at least two occasions recorded more than 12 hours apart. The GW at new onset of hypertension and SPIP were specified in each subject. PE was diagnosed in women that showed both hypertension and SPIP on and after GW 20. Biochemical procedures in the urine and blood specimens. All spot urine samples were coded and processed within 2 hours of collection. Urine samples were transferred into tubes and centrifuged at 700 × g for 5 minutes. Urinary supernatant was stored at -20 °C until measurement of protein, creatinine, and nephrin levels. Protein and creatinine concentrations were measured using a Protein Assay Rapid Kit Wako and Laboassay Creatinine (Wako Pure Chemical Industries, Osaka, Japan), respectively. Nephrin concentrations of all 254 urine samples were measured using a commercial ELISA kit (listed as #1035 in a catalog of Funakoshi Co. Ltd., Tokyo, Japan; human nephrin ELISA kit manufactured by Exocell, Philadelphia, Pennsylvania, USA) that was different from the kit used in our previous study 6 (listed as #1019 in the catalog of Funakoshi Co. Ltd., Tokyo, Japan; nephrin ELISA kit manufactured by Exocell, Philadelphia, Pennsylvania, USA). Urine samples were diluted in 1:10 for nephrin assay. The range of the standard curve is 0.8-200 ng/mL according to the protocol of the human nephrin ELISA kit (listed as #1035 in the catalog of Funakoshi Co. Ltd.). The detection limit of urine nephrin concentration calculated as three times the standard deviation of reagent blank (measured by this kit) was 0.26 ng/ mL. Similarly, the detection limit was 5 μ g/mL for protein. We assumed that samples with undetectable levels contained 0.13 ng/mL nephrin and 2.5 μ g/mL protein. Nephrin and protein concentrations in the urine were corrected by urine creatinine concentration and were expressed as nephrin:creatinine ratio (NCR, ng/mg) and protein:creatinine ratio (PCR, mg/mg). Serum samples collected at the time of urine collection were prepared according to the standard operating procedure and stored at -20 °C until measurement of sFlt-1 and PlGF using commercial ELISA kits (R&D Systems, Minneapolis, MN, USA). Serum samples were diluted in the range of 1:10-1:100 for sFlt-1 assay and 1:1-1:10 for PlGF assay.
Statistical analyses. Data are presented as the median (range) and or mean ± SD. Statistical analyses were performed using the JMP10 © statistical software package (SAS, Cary, NC, USA). The Mann-Whitney U test was used to compare median or mean values between two groups. The Kruskal-Wallis test was used for comparison of medians or means of three groups. Receiver operating characteristic curves (ROC) were constructed for the biomarkers to assess their ability to differentiate women with PE onset later. The Spearman's rank-order correlation was used to test associations between two variables. In all analyses, P < 0.05 was taken to indicate statistical significance. However, a significant finding regarding a linear correlation between two variables was defined as that meeting both P < 0.05 and correlation coefficient (r) > 0.25 or < -0.25.