R statistical language and environment

Normality of continuous variables was assessed using the Kolmogorov-Smirnov test and visual inspection of histograms. The chi-square test was used to evaluate the differences in proportions. Spearman’s rank correlation coefficients were used to determine the relationship between the plasma soluble adhesion molecules and the plasma pro-inflammatory cytokine concentrations. For concentrations below the detection limit, 99% of the lowest detectable concentration across all samples was used. Group-level differences in concentrations were evaluated using the Kruskal-Wallis test, and pairwise comparisons were performed using the Mann-Whitney U test. An alternative analysis was performed using linear models with adjustment for gestational age at sampling. A p-value of <0.05 was considered significant. Analysis was performed using the R statistical language and environment (www.r-project.org).

Biomarker data were log transformed to improve normality and then compared between gestational age-matched groups using two-tailed t-tests. Logistic regression was used to combine the two biomarkers and test the synergistic effect (interaction) between sFas and Elabela for the prediction of either early- or late-preeclampsia. Receiver operating characteristic (ROC) curves were constructed and the area under the ROC (AUC) was calculated for individual biomarkers and the combined model. Analysis was performed using the R statistical language and environment (www.r-project.org), version 3.5.2. To determine the potential effect of covariates on the predictive performance of the combined model, sFas and Elabela concentrations were converted into multiple of means (MoM), accounting for covariates with a significant effect among gestational age, smoking status, parity, and body mass index. For patient demographics data, continuous variables were compared between groups with the Mann-Whitney U tests and the Fisher’s exact test was used for comparisons of categorical variables between groups. Statistical analysis was performed using the SPSS software version 19 (IBM Corporation, Armonk, NY, USA). A p-value <0.05 was considered statistically significant.

Statistical analysis was performed using the R statistical language and environment (www.r-project.org). Data was compared between groups using unpaired Wilcoxon tests, and p-values were adjusted across comparisons and the two analytes (IL-6 and ASC) to control the false discovery rate. Adjustment for gestational age at sampling was performed using a linear regression model. An adjusted p-value (i.e. q-value) <0.05 was considered a significant result. The magnitude of differences was expressed as the difference in the means after log₂ transformation of the data, to obtain log₂ fold changes in the concentrations. The correlation between ASC and IL-6 levels was assessed via Spearman correlation tests and was inspected using locally weighted scatterplot smoothing (LOESS).

Least-square multiple regression was used to examine the effects of ethnicity, maternal age, body mass index (BMI), obstetrical history and cervical length (expressed as a continuous variable) on SWS. Logistic regression was used to examine the effect of continuous variables (BMI, maternal age) on the probability of having a short or a soft cervix. Variance inflation factors (VIF) were estimated to ensure that co-linearity among variables did not affect the reliability of the estimation of the coefficients in the logistic regression model. The effects of parity, prior PTD, cervical length dichotomized as short (≤25 mm) and not short (>25 mm), cervical softness dichotomized as soft (SWS <25th centile for gestational age) and not soft (SWS ≥25^th centile for gestational age), and their interactions on the risk (incidence) of sPTD <34 and <37 weeks were examined by fitting hierarchically nested log linear models to multi-way contingency tables of the data. Relative Risks were estimated using the Poisson regression model with robust estimation of the variance structure. Two-tailed Fisher’s exact tests were used to compare simple differences between proportions and 2 × 2 marginal tables. Values for statistics below the 5% significance level were accepted as statistically significant. The analyses were carried out using the R statistical language and environment (www.r-project.org).

Descriptive statistics for demographic characteristics, disease presentation, and treatment related toxicity were summarized. Continuous variables were analyzed using the Mann-Whitney test or the Kruskal-Wallis test as appropriate, and categorical variables were analyzed using the Fisher’s exact test. Age was analyzed as a continuous variable in our univariate analysis, and as a categorical variable in the OS analysis. PS was analyzed as a categorical variable depicting two groups of PS 0–1 and 2–3. The two-sample, binomial test of proportions was used to compare proportions between different groups. All tests were two sided and utilized a 5% Type I Error. The R statistical language and environment (www.r-project.org) and the STATA package were used for computations¹⁵ . Univariate log-rank tests and Cox proportional hazards models were used to correlate treatment, demographic, and clinical variables with OS. OS was calculated from date of diagnosis of colorectal cancer to date of death. Kaplan-Meier survival curves and hazard ratios (with corresponding 95% confidence intervals) were computed.

Pairwise differences in maternal plasma concentrations of PAPP-A, PAPP-A2, IGFBP-1, and IGFBP-4 in samples collected preterm were assessed by utilizing linear models in which data were log transformed to improve normality. Relevant covariates adjusted for in the analyses were selected using stepwise backward elimination from among the following: gestational age at sampling, sample storage time, body mass index, maternal age, and parity. When comparing term and preterm delivery groups, samples collected before 26^3/7 weeks of gestation were removed to match the distribution of samples. The comparison between the term groups was performed using the same methods. The significance of the group variable coefficient was assessed via t-test, with p<0.05 deemed a significant result. Analysis was performed using the R statistical language and environment (www.r-project.org).