Prognostic factors for cesarean section outcome of pregnant women with gestational diabetes mellitus: a systematic review and meta-analysis

Background

In pregnant women with gestational diabetes mellitus (GDM), the overall cesarean section rate was accounted for 35.3%.¹ Simultaneously, compared with nondiabetic pregnant women, diabetic maternal acute cesarean section rate was reported 1.52 times of GDM.² Diabetes is an important risk factor for surgical incision infection,³ and for cesarean section, diabetes is an important risk factor for maternal postoperative wound infection as well.⁴ Thus, the pregnancy with diabetes and the management of special risk factors are important, and the existing systematic evaluation shows that effective treatment and control of GDM can reduce preeclampsia, shoulder dystocia, and the incidence of huge children.⁵ In addition, several systematic reviews have concentrated on the effects of certain specificc factors based on the health outcomes of pregnant women with GDM, such as the effects of different glycemic management regimens on glycemic control and maternal and child outcomes,^6–10 and effects of dietary intervention or nutritional therapy based on maternal and child outcomes.^11,12 For patients with cesarean section with GDM, there have been several studies evaluating differences in patients’ outcomes under different conditions, such as anesthesia,^13–15 postoperative fluid regimen,^16,17 and postoperative wound care.¹⁸ However, there is no systematic review regarding the current evaluation of the factors affecting the maternal and child’s outcomes during the period of affecting by GDM. This study was designed to assess the risk factors associated with perioperative outcomes in pregnant women with GDM.

Methods

Inclusion and exclusion criteria

Inclusion criteria: 1) pregnant women suffered cesarean section with GDM; 2) exposure factors for cesarean outcomes; 3) RCTs; and 4) reported perioperative outcomes, such as blood glucose level, Apgar scores, adverse effects, and so on. Exclusion criteria: 1) there were no specific outcome data to assess the impact of exposure factors on patients with perioperative outcomes; 2) non-English and Chinese published research, 3) summary of unpublished meeting.

Literature search

We conducted a systematic search on Medline (via PubMed), EMBASE, Cochrane Library, CBM, CNKI and Wanfang, using the terms Diabetes, Gestational, Diabet*, “Cesarean Section”, caesarean, “diabetes, pregnancy”, “gestational diabetes mellitus”, “cesarean section”, “cesarean section”, “caesarean section”. The retrieval date was February 28, 2018.

Study screening

Two researchers independently screened the literature titles, abstracts, and the full text. A pre-test was performed prior to formal screening of the literature to ensure that each researcher truly perceived the screening criteria and process. Discrepancies between the two reviewers were resolved by consensus discussion.

Data extraction

The two researchers independently extracted the following data from the pre-designed information extraction table: year of publication, name of journal, the first author’s affiliation, place and duration of study, funding, conflict of interest, type of study, sample size, basic characteristics of study object, exposure factors, and associated outcome data. A pretest was conducted before formal extraction to ensure that each researcher agrees with the extraction criteria and process. If there are some differences, they could be solved through discussion.

Risk of bias assessment

Two researchers used the Cochrane risk of bias tool was used for bias risk assessment of randomized controlled trials. A pretest was conducted before the formal evaluation to ensure that each researcher agrees with the evaluation criteria and process. In case of existence of some differences, they could be solved by a third researcher.

Data consolidation and analysis

In the RevMan 5.3 software, the RR and 95% CI were used to combine the binary data, and the data were merged using the mean difference (MD) and 95% CI. The data combination uses a random-effect model. The heterogeneity was included in the study by Cochran’s Q test (P<0.05 denotes heterogeneity) and I² test. When the number of inclusion indicators is ≥10, the publication bias is evaluated by making a funnel plot; conversely, the qualitative analysis was included in the study funding, the conflict of interest, and the outcome to discuss the possibility of publication bias.

Quality of evidence

The quality of the evidence was graded according to the principles of the GRADE approach used in the evaluation of prognostic studies^19,20 and in a previous study (as example).²¹ (These factors may lead to rating down the quality of evidence in GRADE system) and the three upgraded factors (large effect, dose-response, and plausible confounders) to determine the final level of evidence. Quality of evidence was ranked as high, medium, low, and very low-level using the results of summary table.

Evidence mapping

Excel 2016 was used to integrate the RR value from meta-analysis and GRADE. The result is shown intuitively with the bubble diagram. Due to heterogeneity of MD for the outcome, we did not make a bubble diagram for MD value from meta-analysis.

Results

Study selection

There are 13,447 articles identified by literature search. After duplicates were removed in endnote, 11,585 records titles and abstractswere reviewed, 142 articles were retrieved full-text reviewing. Finally, a total of 12 randomized controlled trials^{14,16–18,22–29} involving 1,390 patients were included for meta-analysis (Figure 1).

Figure 1 The screening flow chart.

Characteristics of the included studies

The studies were published in 2010 and 2017, the sample sizes ranged from 33 to 201. All studies were from China. The participant age was from 24 to 39. The two studies^16,23 were funded by nonprofit funds, one study reported that there was not the conflict of interest, and the rest of the study did not report funding (Table 1).

Table 1 Basic characteristics of included studies

Risk of bias for included studies

The included RCTs were only low risk of bias in incomplete outcome data and selective reporting (Figure 2); 8 studies^{14,17,18,24–26,29} did not report random sequences; 1 study²⁷ reported that there was a high risk of bias in random sequences; none of the studies reported allocation concealment; 6 studies^{16,18,26–29} did not blind the researchers and patients, and they likely contained an impact on the results; 8 studies^{14,18,22–26,29} did not blind the outcome evaluators, and they likely contained an influence on the results.

Figure 2  Risk of bias graph.

Prognostic factors

Insulin pump

One randomized controlled study²⁵ reported a total of 3 outcomes. It was revealed that duration of treatment process (MD=−5.30, 95% CI: −5.78~−4.82, P<0.00001), insulin dosage (MD=17.00, 95% CI: −23.04~−10.96, P<0.00001), and the incision healing duration (MD=−4.40, 95% CI: −5.58~−3.22, P<0.00001) of the repeated subcutaneous injection for insulin group were superior to those of the insulin pump group, and the difference was statistically significant (Appendix 1).

Short-term fasting and water deprivation

One randomized controlled study²⁷ reported a total of 11 outcomes. Preoperative blood glucose concentrations (MD=0.84, 95% CI: 0.42~1.26, P=0.001) and the level of blood glucose in newborn infants after birth (MD=0.45, 95% CI: −0.03~0.87, P=0.038) for short-term group were superior to those of long-term group, and the difference was statistically significant. Bleeding volume during cesarean section (MD=−42.71, 95% CI:−82.55~−2.86, P=0.039) for short-term group was inferior to long-term group, and the difference was statistically significant. There were no significant differences in postoperative blood sugar concentrations, the rates of nausea and vomiting, incidence of hypoglycemia in newborn infants after birth and mothers before cesarean section, duration of anus exhausting of puerpera, Apgar score 1 and 5 mins after delivery (Appendix 1).

Individual health education

A randomized controlled study²⁸ reported a total of 4 outcomes. Control rates for 2 hr plasma glucose (PG) (RR=1.31, 95% CI: 1.04~1.66, P<0.05) and midnight blood glucose (RR=1.23, 95% CI: 1.01~1.50, P<0.05) and the satisfaction rate of nursing services (MD=6.51, 95% CI: 5.80~7.22, P<0.01) for individualized health education group were superior to those of conventional health education group, and the difference was statistically significant. There were no significant differences in control rates of fasting blood glucose (FBG) as well (Appendix 1).

Fructose injection

Two randomized controlled trials^16,17 reported a total of 8 outcomes. The blood glucose levels 1.5–2 hours after infusion (MD=−1.17, 95% CI: −1.93~−0.41, P=0.003), blood glucose levels 3–4 hours after infusion (MD=−0.99, 95% CI: −1.61−0.36, P=0.002), the level of insulin 1.5 hours after infusion (MD=−13.50, 95% CI: −19.02~−7.98) (P<0.00001), and the level of insulin 3 hours after infusion (MD=−8.59, 95% CI: −13.75~−3.43, P=0.001) for fructose injection were superior to glucose and Insulin injection. The difference was statistically significant. There were no significant differences in blood glucose level, blood glucose level, urinary carcass positive rate, and urine sugar positive rate after transfusion, and no significant difference was found between the two groups (Appendix 1).

Patient-controlled epidural analgesia (PCEA)

A randomized controlled trial²⁴ reported a total of 13 outcomes. The level of blood glucose in presence of analgesia after 6 hours (MD=−0.80, 95% CI: −1.01 to −0.59, P<0.00001), 12 hours (MD=−0.76, 95% CI: −1.00 to −0.52, P<0.00001), 24 hours (MD=−0.65, 95% CI: −0.87 to −0.43, P<0.00001), and 36 hours (MD=−0.75,95% CI: −0.96~−0.54, P<0.00001) for the patient-controlled intravenous analgesia (PCIA) group was superior to PCIA group. The difference was statistically significant. There was no significant difference between the two groups (Appendix 1).

Microwave treatment for postoperative wound

A randomized controlled trial¹⁸ reported a outcome (RR=1.15, 95% CI: 1.03~1.29, P=0.01, see Appendix 1, in which the difference was statistically significant (P<0.01) (Appendix 1).

Psychological intervention (including music therapy)

Two randomized controlled trials^22,23 reported a total of 14 outcome indicators, in addition to entering the operating room immediately with heart rate (MD=−0.86, 95% CI: −2.69~0.97, P=0.36), entering the operating room immediately with anxiety score (MD=−0.13, 95% CI: −2.57~2.31, P=0.92), and the normal feeding rate (RR=1.07, 95% CI: 0.99–1.16, P=0.09). The difference between the two groups was not statistically significant. The rest of the outcome indicators for the psychological intervention group were inferior to the conventional nursing group, and the difference between the two groups was statistically significant (Appendix 1).

Epidural anesthesia

A randomized controlled trial¹⁴ reported a total of six outcomes: glucose concentration at the of cutting skin (MD=1.48, 95% CI: 1.31~1.65, P<0.00001, 2 hours after delivery (MD=0.90, 95% CI: 0.71 to 1.09, P<0.00001) and 6 hours after delivery (MD=1.11, 95% CI: 0.93 to 1.29, P<0.00001), the epidural group were higher than the general anesthesia group, and the difference was statistically significant. The differences in the other outcomes between the two groups were not statistically significant (Appendix 1).

Low-dose sufentanil combined with bupivacaine of spinal-epidural anesthesia

One randomized controlled study²⁶ reported a total of 11 outcomes. Based on the report, glucose concentration at the time of cutting skin (MD=−1.45, 95% CI: −1.61~−129, P<0.00001), the blood glucose concentration 2 hours after delivery (MD=−0.89, 95% CI: −1.07~−071, P<0.00001), and the mean arterial pressure 5 mins after anesthesia (MD=5.80, 95% CI: 3.12~8.48, P<0.00001) were assessed. The difference was statistically significant. However, the differences for the other outcome indicators between the two groups were not statistically significant (Appendix 1).

Addition of once antibiotic

One randomized controlled study²⁹ reported a total of 12 outcomes. The treatment efficiency (RR=0.68, 95% CI: 0.04~1.66, P<0.05) of the addition of an antibiotic once group was inferior to 24 hours antibiotic application group, and the difference was statistically significant. There were no significant differences in response rate, overall response rate, the inefficiency rate, the duration of WBC <12×10⁹/L, body temperature (without fever or returned to normal status 2 hours after surgery), and the grades A, B, and C of healing (Appendix 1).

Publication bias

The number of studies included in every outcome was <10; thus, it was unattempted to use funnel plot to assess the publication bias. All studies have not reported conflict of interest, and only 2 studies reported that the funding originated from the nonprofit grants.

Quality of evidence

The levels of evidence for all the outcome all is low on the GRADE system (see Appendix 2). The reasons for downgrading includes the risk of bias (no randomized sequence generation and allocation concealment, no blindness to researchers, patients and outcome evaluators) and inaccuracy (sample size is less than the optimal sample size and the confidence interval of the combined results cross invalid line).

Evidence mapping

Each bubble corresponds to one outcome for the prognostic factors. The size, color, and position of the bubbles were used to indicate the current research status. The size of the bubbles indicates the sample size, and the color of the bubbles indicates the quality of the evidence. The horizontal coordinate indicates the prognostic factors, the vertical coordinate indicates the RR of meta-analysis (Figure 3).

Figure 3 Summary of risk ratio and quality of evidence of outcomes for the prognostic factors with the bubble diagram.Abbreviation: PCEA, patient-controlled epidural analgesia.

Discussion

The International Federation of Gynecology and Obstetrics (FIGO) guideline recommended to receive cesarean section to prevent shoulder dystocia or birth injury, when fetal weight would be>4,000 g.³⁰ For pregnant women with cesarean section, in addition to the conventional perinatal management, the integration of perioperative management is required, including blood sugar control, anesthesia, healthcare, etc. This study is the first systematic review of the prognostic factors. The results of the systematic review show that the perioperative blood glucose management regimens, preoperative fasting and water-deprivation regimens, anesthesia regimens, postoperative regimens, postoperative analgesia regimens, postoperative wound care regimens, psychological interventions, and different dosing regimens for antibiotics may affect the health outcomes of diabetic maternal and newborns. However, the quality of evidence was low, and more high-quality clinical research evidence is required.

According to the principle of GRADE method in the evaluation of a prognosis research system,^10,11 the quality of evidence for each outcome is low, and the reason of downgrading is mainly bias risk and inaccuracy. The bias risks included in the randomized controlled trials were assessed by the Cochrane Bias Risk Assessment Tool, in which the main source of bias was the nonreported random sequence generation and allocation concealment, which did not blind the researchers, the patients, and the outcome evaluators, and did not report the source of the information and the method of recruiting or joining the patients. The reason for the imprecision is that the sample size is less than the optimal sample size, and the CI of the effect sizes spans the invalid line. For publication bias, as the number of included studies was <10, the publication bias was not evaluated using a funnel plot. In addition, the included studies did not report the conflict of interest, considering the research topics and manufacturers that may be the interests of the relationship, and in addition to psychological intervention and obesity factors, the rest of the comparison groups were assessed by the possibility of publication bias. However, it was not possible to quantify the possibility of publication bias; thus, the publication bias was not considered in this study. In addition, because most of outcomes included only 1 study, a few outcomes included only 2 studies, and I² values are small, it is not been downgraded due to heterogeneity. We performed meta-analyses by using the random effects model for multiple risk factors and outcomes. The qualities of the evidence for all outcomes were low. As the number of studies increases and the quality of the research improves, new research data may change the results of this system review. Therefore, it needs to more new high-quality research to update the review in the future.

The main advantages of this systematic review are: 1) for the first time on the impact of pregnancy in patients with diabetes maternal–perinatal outcome of the perioperative factors were evaluated; 2) the original research carried out a systematic, comprehensive search, greatly reducing the possibility of missing; 3) the quality of evidence was graded by GRADE method, and the factors affecting the outcome of perioperative period of cesarean section in pregnant women with diabetes mellitus were clearly presented and interpreted. The limitations of the system review: 1) Only the studies published in Chinese and English were searched, the other languages were not be considered; 2) All the studies are from China, the results may not been applied to other countries and regions.

Conclusion

Low-quality evidence shows that perioperative blood glucose management regimens, anesthesia regimens, postoperative fluid regimens, postoperative analgesia regimens, postoperative wound care, and psychological interventions may affect the health outcomes of diabetic maternal and newborns.