Strengths and Limitations
This study explored CVF cytokine trajectories throughout spontaneous labor onset and considers symptom status, sub-categories of delivery and time-dependent properties biomarkers. In addition, this study attempts to address gestational age heterogeneity and suggests a pragmatic way that could make diagnosis and prediction of spontaneous labor easier with greater accuracy and at a lower cost. Timely detection of spontaneous labor is fundamental to identify those mothers at higher risk of delivering and plan appropriate management. Conversely, predicting spontaneous delivery to identify women at low risk would avoid the use of unnecessary and sometimes costly interventions.31
Amongst the limitations to be acknowledged, the research design limits the ability to establish real temporal relationships, which might require further confirmation in a longitudinal setting following the same women.
A common characteristic of cytokine data sets is undetectable values or when concentrations exceed a certain threshold on the top of the measurement scale, and thus the issue of “censored data” needs to be addressed, which generally make statistical analysis challenging. The literature recommends different methods to deal with censored data.35 We used simple substitution for IL-1Ra (the only cytokine with censored values) due to the large proportion of censored values. However, simple substitution is not advisable as it may lead to strongly biased results.36 To overcome this, we performed multiple imputation using chained equations to account for variability of missing data; we performed a sensitivity analysis of complete-cases versus multiple imputed data and observed no significant results discrepancies.