Systematic review and meta-analysis of the effectiveness of pre-pregnancy care for women with diabetes for improving maternal and perinatal outcomes

Two reviewers independently assessed the risk of bias for each cohort/ case control study using The Newcastle-Ottawa Scale (NOS) [19]. The criteria assessed for cohort studies were: participants’ selection, comparability of groups and assessment of outcome. While participants’ selection, comparability of groups, and exposure criteria were used to assess the case-control studies. The maximum number of stars awarded for any study were nine: four stars awarded for selection of participants (exposed and non- exposed), ascertainment of exposure and temporal relation between exposure and outcome, two stars were awarded for comparability, if analysis controlled for confounding factors, and three stars were awarded for outcomes if the length of follow up was adequate, with no attrition bias, and if the outcomes were assessed independent of exposure. Studies at “high risk of bias” score less than six stars or scores no stars in comparability domain irrespective of the number of stars scored. Any difference in grading of studies was reconciled by discussion or by involving a third reviewer. For randomized controlled trials, we used the Cochrane tool for bias assessment [20].

We made explicit judgements about whether studies included in the meta-analysis of each of the main outcomes, were at high risk of bias according to the NOS criteria. We assessed the likely magnitude and direction of the bias and whether the likelihood of having an impact on the findings was of any significance.

We assessed the presence of publication bias using Funnel Plots of effect size against standard error for each meta-analysis that included ten or more studies according to Cochrane collaboration methodology. Three analyses were eligible for publication bias assessment including: the effect of PCC on congenital malformations (25 studies), HbA1C (24 studies) and perinatal mortality (ten studies). The vertical axis of the plot represents the standard error, while the horizontal axis represents the logarithmic scale of risk ratio for dichotomous variables in case of congenital malformations and perinatal mortality. In the case of continuous variables, as in HbA1c, the horizontal axis represents the standardized mean difference. Furthermore, we assessed selective reporting in all outcomes [20].

The overall quality and strength of evidence for the main outcomes were assessed using the GRADE approach [21]. We created a 'Summary of findings' tables for the main outcomes of the review. The body of evidence is downgraded from ‘high quality’ by one level for serious (or by two levels for very serious) limitations depending on assessments of risk of bias, indirectness of evidence, inconsistency, imprecision of effect estimates or potential publication bias. Subsequently, the quality of evidence was graded as ‘high’, ‘moderate’, ‘low’ or ‘very low’ certainty. Evidence derived from observational studies receive an initial grade of ‘low’, however, we upgraded the quality of evidence when there was a large magnitude of effect (RR>2 or RR<0.5, in the absence of plausible confounders) [21]. We downgraded scores for risk of bias (weight of studies show risk of bias as assessed by low NOS <6), inconsistency (unexplained heterogeneity), indirectness of evidence (presence of factors that limit the generalizability of the results), imprecision in the pooled risk estimate (the 95% CI for risk estimates are wide or cross a minimally important difference of 10% for benefit or harm (RR 0.9–1.1)), and publication bias (evidence of small-study effects) [21]. We used the GRADEpro tool in order to create the 'Summary of findings' tables [22]. We assessed the quality of the body of evidence relating to the following outcomes for the main comparison, PCC versus no PCC; 1) Congenital malformations 2) HbA1c in the first trimester of pregnancy 3) Perinatal mortality 4) Preterm delivery 5) Maternal hypoglycemia 6) Gestational age at booking for antenatal care.

We produced a summary of the intervention effect using the GRADE approach, a measure of quality for each of the above outcomes.

A statistical analysis using RevMan 5 software (RevMan 2014) was carried out [23]. We used the fixed-effect model to conduct meta-analyses. The pooled statistics was reported as either relative risk (RR) for categorical variables, or mean difference (MD) for continuous variables in the comparison between the intervention and control groups with 95% confidence intervals (CI). Heterogeneity was quantified in each meta-analysis using the Tau², I² and Chi² statistics [24]. We regarded heterogeneity as substantial if I² was ≥ 50% and either Tau² was greater than zero, or there was a low p value (less than 0.10) in the Chi² test for heterogeneity.

We conducted sensitivity analyses by excluding studies with high risk of bias from the metanalysis for the main review outcomes. We conducted sensitivity analysis for two maternal outcomes which are gestation age at the first antenatal visit and first trimester HBA1c level, in addition to five neonatal outcomes, including: congenital malformations, preterm delivery, perinatal mortality, SGA, and admission to NICU.

The authors decided to utilize the modified version of NOS proposed by the Agency for Healthcare Research and Quality (AHRQ), U.S. Department of Health and Human Services [25, 26] as it was more rigorous, robust, and user-friendly than the original version proposed by the University of Ottawa [19].

We conducted additional sensitivity analysis by excluding studies with high risk of bias from certain outcomes

The main characteristics of included studies are summarized in Tables 1–3. There were 36 included studies conducted through March 2019, all of which were conducted in high income countries [35]. Of the included studies 18 were prospective cohort studies [31, 36–52], 16 were retrospective cohort studies [29, 32, 33, 49, 53–65], one was a trial [66], and one was a case control study [67]. The number of participants of cohort studies were 8199 women, among whom 3213 received PCC followed by antenatal care, whilst 4986 only received antenatal care. There were 24 participants in the case control study [67] and 180 pregnancies in the trial [66]. Most of studies did not report the differences in the outcomes among type 1 versus type 2 diabetes, subsequently, we could not conduct the analysis separately for each type of diabetes.

The PPC in all the cohort studies included control and self-monitoring of blood glucose except for one study which was designed to examine the effectiveness of pre-pregnancy counseling on perinatal outcomes [40]. In addition to glycemic control, ten studies included screening and treatment of complications of diabetes in the PPC program [29, 42, 45, 50, 53, 55, 56, 58, 61, 62]. Eleven studies (12 reports) reported comprehensive PCC program including control and self-monitoring of blood glucose in addition to any combination of the following: folic acid supplementation, diet and exercise, smoking cessation, alcohol withdrawal advice, and discontinuation of teratogenic drugs [31, 38, 47, 50–52, 58–61, 63].

In this review, a total of 14 outcomes were reported in the cohort studies, including five maternal outcomes: HbA1c in the first trimester, CS delivery, miscarriage, GA at first antenatal booking, and maternal hypoglycemia during the first trimester. There were nine neonatal outcomes (congenital malformations, perinatal mortality, preterm delivery, macrosomia/LGA, SGA, neonatal hypoglycemia, admission to NICU, RDS and shoulder dystocia). The most frequently reported outcomes were HbA1c in the first trimester (24 studies) [31, 32, 36, 39, 44, 46–57, 60–65, 68] and congenital malformations (25 studies) [29, 31, 32, 36–48, 50–54, 56, 59, 62, 65]. Fourteen studies examined the reduction in CS delivery in women who received PCC compared to those who did not [31, 41, 43, 44, 50–53, 56–59, 63, 65]. Eleven studies compared the miscarriage rate between the two groups [31, 39, 41–43, 50, 51, 55, 59, 63, 65].

Perinatal mortality [31, 40, 41, 43, 45, 50, 52, 56, 59, 62] and preterm delivery [31, 41, 43, 50, 56, 58, 59, 63, 65] were reported in ten and nine studies respectively, while LGA, macrosomia [31, 44, 50–52, 56, 59, 63, 65] SGA [41, 50, 52, 56, 59, 63], and neonatal hypoglycemia [41, 43, 50, 53, 65] were reported in nine, six and five studies respectively. Admission to NICU [50–52, 56], and neonatal RDS [41, 43, 53, 65] were reported in four studies. The least reported outcomes were maternal hypoglycemia in three studies (all reported results among type 1 diabetes) [31, 33, 51] and shoulder dystocia in two studies [50, 65].

We used the NOS Form for Cohort Studies to determine the level of bias of cohort studies included in this review (Table 4). 21 studies were determined to be at a low risk of bias [30–33, 41, 43, 45, 49–52, 55, 57–65], while 15 studies were judged to be at a high risk of bias [27–29, 36–40, 42, 44, 46–48, 53, 54, 56, 68]. Some of the studies at a high risk of bias were initially designed to assess aspects of PPC other than its effectiveness in improving maternal and perinatal outcomes, hence the poor methodological design when assessed with the NOS [40, 42, 56].

The cohort studies included in this review (Table 1) had adequate description of participants including comparison between the PPC group and the control group regarding some confounding factors such as the duration of diabetes and frequency of renal and vascular complications. However, most studies did not address the effect of the confounding factors on the outcomes except for ten studies (11 reports) which used regression analysis to evaluate the independent effect of the PPC [31, 45, 51, 52, 57–60, 63, 65]. In most cohort studies, blinding of the control group was adequate because they were recruited after the inception of pregnancy while attending the antenatal care. In four studies, [33, 38, 49, 50] blinding was inadequate as the control groups were invited to participate in the PCC program and hence were informed about it. All participants received the same antenatal and postnatal care except for six studies [42, 47, 49–52] where participants were followed up in different health settings. In all cohort studies, the compliance of participants to follow up was adequate except for Rosenn et al [39] where 52% of the PCC group were lost to follow-up, and Jensen et al [68] where 3 out of 11 participants in control group did not comply with study protocol as they rejected the self-glucose monitoring. The assessors of the outcomes were not blinded to the participants' allocation; however, we do not believe this would have introduced bias due to the objective nature of the outcomes in this review.

One case control study was included in this review [67]. The study encompassed a small sample size which included 12 cases each for cases and control. Both recall bias and detection bias cannot be excluded.

One trial was included in this review [66] (Table 3). The design of the trial was not clear; neither the method of randomization nor the allocation concealment was described. In addition to that, lack of blinding introduced bias because both groups were aware of the importance of the glycemic control and the complications of diabetes during pregnancy.

Fourteen outcomes were identified after examining all the studies; meta-analysis was possible for 34 cohort studies with 8199 participants.

Gestational age at booking for antenatal care. The results of the meta-analysis on the effect of PCC on the early booking for antenatal care showed that women who attended PCC booked approximately ten days earlier for antenatal care (MD 1.31; 95% CI: 1.40–1.23); five studies, 1081 women very low-certainty evidence)) (Fig 2) (Table 5). The quality of evidence was downgraded from low-grade (observational study) to very low-grade due to the high risk of bias in the study with the largest weight [39] and high unexplained heterogeneity (Table 6).

Congenital malformations. The result of the meta-analysis on the effect of PCC on congenital malformations suggested that PCC resulted in a large reduction in congenital malformations (RR 0.29; 95% CI: 0.21–0.40, 25 studies; 5903 women; high-certainty evidence) (Fig 3) (Table 5). We considered the body of evidence for the effect of PCC on the reduction of congenital malformations to be high quality mainly due to the large effect size with precise and narrow confidence intervals, consistency of direction of effect throughout most of the included studies, no indirectness of evidence, and no heterogeneity or publication bias. Bias in the included studies was considered moderate (59% of the participants were from studies at low-risk of bias) (Fig 3, Table 6).

HbA1c. Meta-analysis of 24 studies which reported HbA1c showed that PCC likely results in a reduction in HbA1c in the first trimester of pregnancy by an average of 1.27%; (MD 1.27; 95% CI: 1.33–1.22; 4927 women; moderate-certainty evidence) (Fig 4) (Tables 5 & 6). We considered bias in the included studies low (77% of the participants were from studies at low risk of bias) (Fig 4, Table 6), while heterogeneity can be explained by long span of time between the first and the last study (1982 and 2017), during which time many innovations in the management of diabetes has occurred with substantial reduction in the target level of HbA1c. The apparent publication bias in this outcome can be explained with the heterogeneity associated with this analysis.

Maternal hypoglycemia. We are uncertain about the effect of PCC on maternal hypoglycemia during the first trimester of pregnancy; (RR 1.38; 95% CI: 1.07–1.79); three studies; 686 women; very low-certainty evidence) (Fig 5) (Table 5). The grade of evidence was downgraded from low to very low due to inconsistency of the direction of effect and high heterogeneity (I² = 76%) in the included studies (Table 6). The true effect is likely to be substantially different from the effect estimated in this review.

Preterm delivery. The results of the meta-analysis on the effect of PCC on preterm delivery indicate that PCC lead to a slight reduction in preterm delivery rate among women with diabetes (RR 0.85; 95% CI: 0.73–0.99; nine studies, 2414 women; moderate-certainty evidence) (Fig 6) (Table 5). We upgraded the body of evidence for the effect of PCC on the reduction of preterm delivery to moderate quality. This upgrade was based on the narrow confidence intervals around the point estimate, consistency of direction of effect in most of the included studies, no indirectness of evidence, low risk of bias of the body of evidence as only 1.9% of the participants were from one study with high risk of bias. The low heterogeneity with no evidence of selective reporting increase our confidence in the outcome of a small reduction in preterm delivery (Fig 6, Table 6)

Perinatal mortality. The meta-analysis results on the effect of PCC on perinatal mortality indicates that PCC results in a large reduction in perinatal mortality (RR 0.46; 95% CI: 0.30–0.73; ten studies; 3071 women; moderate-certainty evidence) (Fig 7) (Table 5). The quality of evidence has been upgraded to moderate due to the narrow confidence intervals, consistency of direction of effect in most of the included studies, no indirectness of evidence, and low risk of bias of the body of evidence as only 7.6% of the participants were from two studies at high risk of bias and no heterogeneity or publication bias.

Small for gestational age. The result of the meta-analysis indicates that PCC may result in large reduction in SGA (RR 0.52; 95% CI: 0.37–0.75; six studies, 2261 women; moderate-certainty evidence) (Fig 8) (Table 5). We upgraded the body of evidence for the effect of PCC on the reduction of SGA to moderate-quality on account of the large effect size (48% reduction in SGA) with precise narrow confidence interval, consistency of direction of effect throughout the included studies, no indirectness of evidence, and no heterogeneity and no evidence of selective reporting. We considered bias in the included studies as low (2% of the participants were from one study at high risk of bias) (Fig 8, Table 6).

NICU admission. The result of the meta-analysis on the effect of PCC on admission to NICU indicates that PCC may reduce the rate of NICU admissions (RR 0.75; 95% CI: 0.67–0.84; four studies; 1322 women; moderate-certainty evidence) (Fig 9) (Table 5). The body of evidence was upgraded owing to precise narrow confidence intervals, consistency of direction of effect, no indirectness of evidence, and no heterogeneity and no evidence of selective reporting. We considered bias in the included studies as low.

Meta-analysis showed that the PCC may have little or no effect in reducing the CS rate (RR 1.02; 95% CI: 0.96–1.07; 14 studies; 3641 women; low-certainty evidence), miscarriage rate (RR 0.86; 95% CI: 0.70–1.06; 11 studies; 2698 women; low- certainty evidence), macrosomia rate, (RR 1.06; 95% CI: 0.97–1.15; nine studies; 2787 women, low- certainty evidence), neonatal hypoglycemia (RR 0.93; 95% CI: 0.74–1.18; five studies; 880 women; low- certainty evidence), RDS (RR 0.78; 95% CI: 0.47–1.29; four studies; 466 women; very low- certainty evidence) and shoulder dystocia (RR 0.28; 95% CI: 0.07–1.12; 2 studies; 530 women; very low- certainty evidence).

We performed sensitivity analysis by excluding studies with high risk of bias from the meta-analysis. Overall, results and conclusions were not changed (Figs 1–7, S3 File).

Publication bias. We examined the possibility of publication bias by evaluating the asymmetry of the Funnel Plots (Fig 10). Analysis of the effect of PCC on congenital malformations and perinatal mortality (Fig 10A & 10B) demonstrated symmetrical distribution of the studies which can reasonably exclude publication bias. Analysis of the effect on HbA1C, showed asymmetrical distribution of the studies (Fig 10C), however this can be explained by the marked heterogeneity associated with this outcome [69]. We found no evidence of selective reporting of outcomes in all included studies.

New strategies for incorporating PCC in ongoing healthcare services, such as adults’ diabetic clinics and primary healthcare may prove to be cost effective and improve the feasibility and applicability of PCC globally.

Incorporation of health education about contraception, fertility of women with diabetes, and the importance of pregnancy planning in the services of diabetic clinics may improve the uptake of PCC.

Further research in interventions for improving pregnancy planning and increase utilization of PCC in different communities may improve our understanding of the poor utilization of PCC and suggest areas for improvement. In addition, research to investigate important outcomes, which are still surrounded by uncertainty, such as the association between PCC and maternal hypoglycemia, should be encouraged. There is lack of studies addressing these problems in low-income countries that raise the need for future research both quantitative and qualitative.