An important strength of our study

An important strength of our study is the use of an elastic net regression model to select the variable and construct the model, which performed ideally in terms of both predictive accuracy and sparsity for the high-dimensional datasets [16]. As a result, the apparent predictive effect of multiple cytokine combinations was determined. In addition, maternal cytokines confer racial difference in the risk of sPTB. To our knowledge, our study was the first (in Chinese) to investigate the relationship between multiple cytokines and sPTB risk. Further studies on gene regulation may help to elucidate the ethnic disparity of sPTB [44], [45]. This study also has some limitations. First, the measurements of some cytokines such as IL-2, IL-10, IL-15, GM-CSF, and MCP-1 were below the LOD for 40% of all the samples, which can be attributed to degradation. We have only compared the samples from the cases and controls from whom blood was drawn at about the same time, and the samples were stored under the same conditions for the same time. Thus, it TAE226 is reasonable to assume that the level of degradation of the cytokine is the same. Moreover, the results were reliable as the cytokine levels were categorized into quartiles for analysis, and hence relative (but not absolute) concentrations were compared. It has been assumed that the rate of change in concentration does not differ between cases and controls, and therefore the resulting misclassification bias will probably be non-differential and hence not influence the observed associations. Second, although we have noted the importance of the gestational age at sampling, the collection of blood samples at multiple time points during pregnancy to examine the longitudinal changes in cytokine levels may not be feasible for an exploratory study. Third, our analysis failed to control the possible influence of air pollution, as air pollution might have influenced the serum cytokine expression through systemic inflammation [46]. In addition, we did not exclude the five women who were taken anti-inflammatory drugs or steroids for three months immediately before inclusion in the study, as the effect of medication on serum cytokine levels might be limited after a certain time. We have also adjusted the potential confounder in our analysis. Finally, the sample size was relatively small. To overcome these shortcomings, we performed 10-fold cross-validation, for the assessment of the stability of the models to evaluate the potential generalization of these results to other datasets and also as a final forward classification step to avoid the overfitting of the predictive models. Even though our findings were internally cross-validated, it is crucial to independently validate in different cohorts.
Conclusion