Study characteristics

The process of literature search in this study is outlined in Fig. 1. We did our initial search in PubMed databases and identified 1171 records. These records were imported into EndNote for further screening. There were no duplicated records among these records, and the review paper (n = 218) and meta-analysis (n = 11) were excluded by keyword and note screening. The remaining 942 records underwent further title and abstract screening, and 908 records were removed because of inappropriate study subjects (n = 112) and research purposes (n = 796). The full text of the remaining 34 studies was carefully reviewed, and 17 studies were excluded because of inappropriate study design (n = 11) or lack of adequate data (n = 6). Finally, 17 studies that met the inclusion criteria were included in the quantitative synthesis.

Fig. 1

The flow diagram of this study selection process in the meta-analysis

Table 1 summarizes the detailed information for the remaining 17 articles. The country, sample origin, gestational weeks for sampling, delivery gestational weeks, significant differences, inflammatory factors, and function of the inflammatory factors, etc. were included.

Table 1 The detailed information for the included articles.Study quality

The qualities of these 17 studies were analyzed in Fig. 2. For the selection bias, 12 studies described the specific methods of randomization. Seven studies on performance bias indicated that double-blind (participants and personnel) was used in these studies. Although the detailed methods of blinding for outcome assessment were not mentioned, the objective measurement of the inflammatory factors is related to double-blinding of participants and personnel. Therefore, the detection of bias was related to performance bias. For attrition bias, 17 studies have complete outcome data. For reporting bias, all the studies have no selective experimental data for high risk. For other bias, 9 studies were judged to have a high risk, including those with a small sample size. Overall, about 71% studies have low risk of selection bias, and 29% studies have unclear risk of selection bias. About 41% studies have low risk of performance bias, and 51% studies have unclear risk of performance bias. About 29% studies have low risk of detection bias, and 71% studies have unclear risk of detection bias. All studies have low risk of attrition bias and reporting bias. About 47% studies have low risk of other bias, and 53% studies have high risk of other bias.

Fig. 2

The risk-of-bias assessments of the included studies. The risk-of-bias assessments of the included studies. Random sequence generation (selection bias), allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias), and other bias were assessed by Review Manager 5.4 as “low,” “high,” or “unclear” risk. Green color indicated low risk of bias, yellow color indicated unclear risk of bias, and red color indicated high risk of bias. About 71% studies have low risk of selection bias, and 29% studies have unclear risk of selection bias. About 41% studies have low risk of performance bias, and 51% studies have unclear risk of performance bias. About 29% studies have low risk of detection bias, and 71% studies have unclear risk of detection bias. All studies have low risk of attrition bias and reporting bias. About 47% studies have low risk of other bias, and 53% studies have high risk of other bias

Main efficacy of meta-analysisThe prioritization of the inflammatory factors in infectious PTB prediction is sTNFR2 > TNFα > IL-10 > IL-6 > CRP > IL-1β

Among the included 17 studies, 1340 normal and 502 infectious PTB pregnancies were analyzed, and 19 inflammatory factors were studied in the peripheral blood of pregnant women. In the peripheral blood of normal and PTB patients, C-reactive protein (CRP), interleukins (IL-1β, IL-2, IL-4, IL-6, IL-8, IL-10, IL-17A, IL-27p28), tumor necrosis factor (TNF)/nerve growth factor (NGF) cytokine family (TNFα, sTNFR1, sTNFR2, NGF), macrophage related soluble factors (sCD14, sCD163), interferons (IFN-γ), chemokines (CXCL9), ferritin, and intestinal fatty acid binding protein (I-FABP) were analyzed. Among 19 inflammatory factors, 6 inflammatory factors were studied more than twice, and IL-1β, IL-6, IL-10, CRP, TNFα, and sTNFR2 were included. We used network meta-analysis for these 6 inflammatory factors to efficiently rank the priority for infectious PTB prediction. The network and cumulative probabilities of these 6 inflammatory biomarkers were shown in Fig. 3A and Figure 3B respectively. The network analyzed results showed in Table 2 indicated that sTNFR2 had the highest SUCRA value (94%), TNFα had a 64.2% SUCRA value, IL-10 had a 63.9% SUCRA value, IL-6 had a 42.5% SUCRA value, CRP had a 18.9% SUCRA value, and IL-1β had the lowest SUCRA value (16.5%). Thus, the prioritization of the inflammatory factors in infectious PTB prediction is sTNFR2 > TNFα > IL-10 > IL-6 > CRP > IL-1β.

Fig. 3

The network and cumulative probabilities of IL-1β, IL-6, IL-10, CRP, TNFα, and sTNFR2. a The network of IL-1β, IL-6, IL-10, CRP, TNFα, and sTNFR2. The size of the nodes and the thickness of the edges are weighted according to the number of studies evaluating each inflammatory factors and direct comparison, respectively. b The cumulative probabilities of IL-1β, IL-6, IL-10, CRP, TNFα, and sTNFR2 to rank the priority for PTB prediction. The curve of each inflammatory factors is the cumulative ranking curve lining out the SUCRA value. (IL-1β indicated interleukin-1 beta, IL-6 indicated interleukin-6, IL-10 indicated interleukin-10, CRP indicated C-reactive protein, TNFα indicated tumor necrosis factor α, and sTNFR2 indicated soluble tumor necrosis factor receptor 2, SUCRA indicated surface under the cumulative ranking curve)

Table 2 The SUCRA values of inflammatory factors for PTB prediction.CRP/IL-1β/IL-6 is reliable cluster for predicting the occurrence of infectious PTB in gestational 27–34 weeks

The studies (Skrablin, S. 2007; Zhu, H. 2018) found that CRP, IL-1β and IL-6 these three inflammatory factors both upregulated in maternal blood of PTB patients in gestational 27–34 weeks comparing with normal [23, 24], and these two studies were included for further analysis. Based on this, the global consideration of multiple inflammatory factors at specific time for infectious PTB prediction is necessary. The forest plot indicated that CRP, IL-1β, and IL-6 all significantly higher in the maternal blood of PTB patients compared with normal (SMD: 3.633; 95% CI: 1.971 to 5.296; P < 0.05, and I2 = 93.5%) in gestational 27–34 weeks (Fig. 4). The funnel plot with pseudo 95% confidence limits was shown in Supplementary data 2A, and the sensitivity analysis shown in Supplementary data 2B did not significantly affect the overall results, supporting the function of CRP, IL-1β and IL-6 together as a reliable cluster for infectious PTB in gestational 27–34 weeks.

Fig. 4

The forest plot of comparison of maternal CRP/IL-1β/IL-6 cluster between normal and PTB patients in gestational 27–34 weeks. SMD: 3.633; 95% CI: 1.971 to 5.296; P < 0.05, and I2 = 93.5%. (SMD indicated standard mean difference; CI indicated confidence interval)

TNF/NGF family is potential cluster for predicting the occurrence of infectious PTB in gestational 25–33 weeks

Signals emanating from receptors of the TNF/NGF family take part in immune defense [25]. TNF, NGF, TNF receptor, and NGF receptor all belong to the TNF/NGF family. Among 17 selected references in this study, 3 references (Escobar-Arregoces, F. 2018; Wallenstein, M.B. 2016; McDonald, C. R. 2019) that related to the TNF/NGF family and analyzed maternal blood in gestational 25–33 weeks of PTB were chosen for further analysis [21, 26, 27]. Based on these, we looked at TNFα, TNFβ, NGF, sTNFR1, and sTNFR2 in these 3 studies to see if the TNF/NGF family participated in the occurrence of infectious PTB. The forest plot result showed that the TNF/NGF family participated in infectious PTB occurrence (SMD: 1.592; 95% CI: 0.020 to 3.163; P < 0.05, and I2 = 98.4%) (Figure 5). The funnel plot with pseudo 95% confidence limits was shown in Supplementary data 3A. To show the impact of each study, we did the sensitivity analysis in supplementary data 3B. The results of sensitivity analysis showed that the three studies all contributed importantly, and excluding either study would affect the statistical difference.

Fig. 5

The forest plot of comparison of maternal TNF/NGF family related factors between normal and PTB patients in 25–33 weeks. SMD: 1.592; 95% CI: 0.020 to 3.163; P < 0.05, and I2 = 98.4%. (SMD indicated standard mean difference; CI indicated confidence interval)