Effectiveness of exercise intervention during pregnancy on high-risk women for gestational diabetes mellitus prevention: A meta-analysis of published RCTs.

OBJECTIVE: We aimed at investigating the preventive role of exercise intervention during pregnancy, in high-risk women for gestational diabetes mellitus (GDM).
MATERIALS AND METHODS: We searched PubMed, CENTRAL, and Scopus for randomized controlled trials (RCTs) that evaluated exercise interventions during pregnancy on women at high risk for GDM. Data were combined with random effects models. Between study heterogeneity (Cochran's Q statistic) and the extent of study effects variability [I2 with 95% confidence interval (CI)] were estimated. Sensitivity analyses examined the effect of population, intervention, and study characteristics. We also evaluated the potential for publication bias.
RESULTS: Among the 1,508 high-risk women who were analyzed in 9 RCTs, 374 (24.8%) [160 (21.4%) in intervention, and 214 (28.1%) in control group] developed GDM. Women who received exercise intervention during pregnancy were less likely to develop GDM compared to those who followed the standard prenatal care (OR 0.70, 95%CI 0.52, 0.93; P-value 0.02) [Q 10.08, P-value 0.26; I2 21% (95%CI 0, 62%]. Studies with low attrition bias also showed a similar result (OR 0.70, 95%CI 0.51, 0.97; P-value 0.03). A protective effect was also supported when analysis was limited to studies including women with low education level (OR 0.55; 95%CI 0.40, 0.74; P-value 0.0001); studies with exercise intervention duration more than 20 weeks (OR 0.54; 95%CI 0.40, 0.74; P-value 0.0007); and studies with a motivation component in the intervention (OR 0.69, 95%CI 0.50, 0.96; P-value 0.03). We could not exclude large variability in study effects because the upper limit of I2 confidence interval was higher than 50% for all analyses. There was no conclusive evidence for small study effects (P-value 0.31).
CONCLUSIONS: Our study might support a protective effect of exercise intervention during pregnancy for high-risk women to prevent GDM. The protective result should be corroborated by large, high quality RCTs.

Introduction

Gestational diabetes mellitus (GDM) is a multifactorial disorder from the interaction between genetic and environmental risk factors. It is characterized by insulin resistance and decreased pancreatic b-cell function. It is also a risk factor for the future development of type 2 diabetes mellitus [1], and one of the most common diseases during pregnancy [2]. The worldwide prevalence is increasing ranging between 2 and 14% [3]. Women with GDM have an increased risk of obstetric, fetal, neonatal, maternal, and child complications [3-11].

Identified risk factors for GDM include obesity [1, 4–6, 9–13], sedentary lifestyle [10], unbalanced diet [3, 10], socioeconomic factors including low education level [14], ethnicity [2, 4, 10, 15] and family history [4, 10]. Besides interventions for treatment [3], several preventive interventions such as behavioral and lifestyle modifications were evaluated [3, 16].

Previous studies on the effect of exercise intervention were conflicting [4, 5, 13, 17]. Several systematic reviews and meta-analyses [4, 6, 7, 9, 17] showed a significant risk reduction among women in the general population while other studies [8, 10, 18] failed to support risk reduction for GDM. Two recent meta-analyses explored the role of exercise on GDM prevention among high-risk women. One meta-analysis [19], showed no benefit of the interventions, including exercise, compared to placebo; while the other [20] supported a significant GDM risk reduction with exercise during pregnancy among overweight and obese women. However, to our knowledge, there was no systematic approach to evaluate exercise as a single intervention during pregnancy on GDM prevention among high-risk women with any of the risk factors for GDM, and who already received standard prenatal care.

Our study aimed at systematically appraise RCTs that assessed the effectiveness of exercise during pregnancy on the prevention of GDM. We included RCTs on high-risk pregnant women with one or multiple risk factors, which compared exercise to standard prenatal care. We performed meta-analysis with special emphasis on issues of potential biases, and sources of study heterogeneity including both clinical and methodological factors that may account for potential variability in study effects.

Materials and methods

Our study was registered in the Open Science Framework (OSF) (Registration DOI 10.17605/OSF.IO/23NJS, https://archive.org/details/osf-registrations-23njs-v1). This systematic review was performed according to PRISMA extension for complex interventions guideline [21].

Search strategy

We searched PubMed, Cochrane Library Central Register of Controlled Trials (CENTRAL), and Scopus (from inception to May 2022). For Pubmed, we used a search strategy including keywords related to exercise, physical activity, and GDM combined with the Cochrane Collaboration search algorithm for RCTs. We conducted a systematic search on Scopus using the same keywords after excluding articles registered in Pubmed. Finally, we searched CENTRAL including the same keywords related to exercise, physical activity, and GDM. Search algorithms were described in detail in S1 Table. Electronic searches were supplemented by perusal of the references of the retrieved papers as well as the references of review articles. One investigator (GIT) screened all databases. For items considered potentially eligible or unclear, after screening the title and/or abstract, the full text was retrieved. A second investigator (AT) checked on the items that the first investigator (GIT) could not decide. Discrepancies were resolved through consensus. For trials that we could not reach a final decision, or the full text could not be retrieved, we contacted investigators when an e-mail address was available. Two consecutive reminders were also sent to non-responders.

Eligibility criteria

We selected trials according to PICO (population, intervention, comparator, and outcome) approach. We accepted randomized controlled trials (RCTs) in English that recruited pregnant women at high risk for GDM. Factors that increased pregnant women’s risk included at least one of the following: increased BMI [1, 4–6, 9–13], sedentary lifestyle [10], family history [4, 10, 22], previous macrosomia [22], unbalanced diet [3, 10], previous GDM [22], non-white ethnicities [2, 4, 10, 14, 22] and age > 25 years [22]. We considered as eligible trials that assessed interventions of any type of exercise during pregnancy. We accepted trials if women in the comparator group received the standard antenatal care. We considered as eligible trials that reported as outcome the onset of GDM. We accepted all modalities for GDM diagnosis. In case of multiple publications of an RCT with results in different follow-up periods, we accepted the publication including the largest sample. We excluded RCTs that were published at the protocol stage, pilot, or feasibility studies, abstracts from conference proceedings, and RCTs that did not report results on the eligible outcome.

Data extraction

Two independent researchers (GIT and KP) extracted the data. Discrepancies were resolved with consensus, and the participation of a third arbitrator (AT) where necessary. The Cohen kappa coefficient with 95% confidence interval (95% CI) was used to evaluate the agreement between the two investigators who independently extracted the data.

Extracted items included the name of first author, year of publication, country, whether the study was a cluster RCT, number of participating centers, study duration, drop-out rate, sample size, factors related to high risk for GDM in the participants, women’s mean age, and number of participating women with low level of education if reported. We also recorded the type of intervention and the care that women in the comparator group received. For assessing the completeness of exercise intervention reporting, we used the CERT (Consensus on Exercise Reporting Template) tool for complex interventions [23]. CERT was proposed to improve reporting of exercise intervention programs in clinical trials. It included 16 items allocated in 7 categories, i.e., materials, provider, delivery, location, dosage, tailoring, and to what extent the exercise intervention was delivered and performed as planned [23]. In addition, we extracted potential side-effects /adverse events that were reported for intervention, and comparator arm. Finally, we recorded the number of GDM events as the outcome, separately in the experimental and the control arm. We also captured information on the method used in each study for the diagnosis of GDM.

Quality assessment of the studies and rating of overall evidence

We used the risk of bias tool proposed by the Cochrane Collaboration [24] for quality assessment of eligible RCTs. Two independent researchers (GIT and KP) extracted the data on quality assessment. Discrepancies were resolved with consensus, and the participation of a third arbitrator (AT) where necessary. In addition, we used the Grading of Recommendations, Assessment, Development and Evaluation tool (GRADE) for rating the overall evidence [25] (GRADEpro, Version 3.6.1. McMaster University, 2011)”.

Statistical analysis

To combine the events of GDM, we performed both fixed effects and random effects model (REM) meta-analyses. In case that large heterogeneity could not be excluded, we reported the REM results (odds ratio with 95% CI) [26]. Heterogeneity was evaluated with Cochran’s Q statistic (statistically significant for P < 0.10); and it was quantified with the I2 metric (low, moderate, large, very large for values of <25, 25–49, 50–74, >75%, respectively) [27]. The main analyses included all available data. We performed separate analyses limited to studies where increased BMI was included in as a risk factor for GDM, and studies that did not consider BMI; studies where the percentage of participating women with low level education was more than 5%; studies that evaluated an intervention delivered individually, and studies that evaluated an intervention delivered in a group; trials that included a motivation component in the intervention, and trials that did not include motivation; studies with an intervention duration more than 20 weeks, and studies with an intervention duration up to 20 weeks. We also performed meta-regression analyses on GDM OR. The effect of baseline risk, and study duration were included individually as covariates in the meta-regressions. For each meta-regression, the slope coefficient with the standard error (SE), the permutation-based P-value (as suggested by Higgins and Thompson [28] and the tau2 were reported. Publication bias was evaluated via the visual analysis of funnel plot, showing a symmetrical inverted funnel in the absence of bias [29]. To further investigate potential asymmetry due to publication bias, we performed the statistical Egger’s test [30]. We also performed separate analyses for studies with low detection bias (studies reporting blinding of outcome assessors); and for studies with low attrition bias (studies with less than 20% of participants lost in follow-up). The level of significance for all analyses, except for Cochran’s Q statistic, was set at P-value < 0.05. For our analyses, we used SPSS 22.0 (SPSS, Inc., Chicago, Illinois, USA), Stata Statistical Software 10.1 (Stata, College Station, TX, USA), and Review Manager 5.4 (Cochrane Collaboration, UK).

Results

Eligible studies

Our search yielded 1566 items (582 in PubMed, 290 in Scopus, and 694 in CENTRAL). We excluded 268 as duplicated. Out of the 1298 remaining items, we excluded 1260 as non-relevant based on the title, or abstract. Thus, we retrieved 37 papers in full text. Out of the 38 articles, we excluded 29; one paper reported a pilot study; 8 studies did not include an eligible population; 7 studies included a non-eligible intervention; and 13 trials did not report the onset of GDM as an outcome. Finally, we included 9 published RCTs as eligible for our study (Fig 1).

Fig 1

Flow chart of study selection process.

Characteristics of eligible studies

Eligible studies were published from 2012 to 2017. Four RCTs [13, 31–33] were conducted in Europe (one study in Netherlands, one in Spain, one in Norway, and another one in Ireland); two studies was conducted in Oceania [34, 35], (one study in New Zealand, and one in Australia), two in USA [2, 36], and another one [5] in China (Table 1). All trials used the participant as the randomization unit and had a parallel design. One study [31] was multi-centered (five participating centers). The duration of the trials ranged from 19 to 60 months. The drop-out rate was < 20% for all studies, except for one study [36] that was 31.9% (Table 1).

Table 1

Characteristics of eligible studies.

First author, publication year	Country	Number of participated centers	Study duration, mo	Drop-out rate n (%)
Oostdam, 2012	Netherlands	5	48	22 (18.2)
Price, 2012	USA	1	45	29 (31.9)
Barakat, 2013	Spain	1	40	82 (16)
Nobles, 2015	USA	1	60	39 (13.4)
Seneviratne, 2015	New Zealand	1	19	1 (1.3)
Guelfi, 2016	Australia	1	37	3 (1.7)
Krohn Garnæs, 2016	Norway	1	22	17 (18.7)
Wang, 2017	China	1	20	35 (11.7)
Daly, 2017	Ireland	1	41	2 (2.3)

mo, months

A total of 1,738 (866 in intervention, and 872 in control group) high-risk women for GDM participated in the eligible trials. Six studies included overweight, and obesity as risk factors [2, 5, 13, 31, 32, 34]. Additional risk factors included history of GDM in three studies [2, 31, 35], history of type 1 and 2 diabetes mellitus in first- and second-degree relatives in two studies [2, 31], history of macrosomia in one study [31], and previously sedentary lifestyle in two studies [33, 36] (Table 2). Mean age ranged from 24.9 to 37.7 years for women in the intervention group, and from 20.3 to 37.7 years for women in the control group (Table 2). One study [2] reported only the age range (18 to 40 years) (Table 2). Percentage of women with low education level ranged from 2% to 34% in the intervention group, and from 7% to 34.7% in the control group. Four studies [32, 34–36] did not report data on participants’ education level (Table 2).

Table 2

Characteristics of participating women in the eligible studies.

First author, publication year	Sample size (intervention / control)	Risk factors for GDM	Mean age (SD), yr intervention / control	Low education level, n (%) intervention / control
Oostdam, 2012	121 (62 / 59)	Obese (body mass index, BMI ≥ 30) or overweight (BMI ≥ 25) AND at least one of the three following characteristics: (1) history of macrosomia (offspring with a birthweight above the 97th percentile of gestational age); (2) history of GDM; or (3) first-grade relative with T2D	30.8 (5.2) / 30.1 (4.5)	16 (34) / 17 (34.7)
Price, 2012	91 (43 / 48)	Previously sedentary women; no aerobic exercise more than once per week for at least the past 6 months	30.5 (5) / 27.6 (7.3)	ND
Barakat, 2013	510 (255 / 255)	Previously sedentary women; not exercising more than 20 min on more than 3 days/week	31 (3) / 31 (4)	54 (25.7) / 75 (34.4)
Nobles, 2015	290 (143 / 147)	Overweight or obese (pre-pregnancy BMI ≥ 25 kg/m2) with a family history of DM or a diagnosis of GD in prior pregnancy, defined according to the ADA criteria	Range 18–40	26 (22) / 31 (27)
Seneviratne, 2015	75 (38 / 37)	Pre-pregnancy BMI ≥ 25 kg/m2	ND	ND
Guelfi, 2016	172 (85 / 87)	Pregnant women with a history of GDM in a previous pregnancy	33.6 (4.1) / 33.8 (3.9)	ND
Krohn Garnæs, 2016	91 (46 / 45)	Pre-pregnancy BMI ≥ 28 kg/m2	31.3 (3.8) / 31.4 (4.7)	1 (2) / 3 (7)
Wang, 2017	300 (150 / 150)	Pre-pregnancy BMI ≥ 24 kg/m2	32.1 (4.6) / 32.5 (4.9)	31 (21) / 40 (27)
Daly, 2017	88 (44 / 44)	BMIs at their first prenatal visit of 30 or greater	30.0 (5.1) / 29.4 (4.8)	ND

GDM, gestational diabetes mellitus; SD, standard deviation; yr, years; BMI, body mass index; T2D, type 2 diabetes; DM, diabetes mellitus; ADA, American Diabetes Association; ND, no data

Interventions evaluated several exercise programs with the use of various equipment (Table 3). Five [2, 13, 31–33] out of the nine studies also included a motivation component in the intervention. Providers included physiotherapists in two studies [13, 31], health educators in one trial [2], exercise physiologist in two RCTs [34, 35], researchers in three trials [5, 32, 36], and fitness specialist with the assistance of an obstetrician in one study [33]. Three trials [31, 34, 35] delivered the intervention individually; one [36] both in group and individually; three studies [13, 32, 33] delivered the intervention in group; and two studies did not report relevant data [2, 5] (Table 3). Seven trials [5, 13, 31–33, 35, 36] evaluated a supervised intervention (Table 3). The duration of the intervention was more than 20 weeks in five trials [5, 13, 31, 33, 36] (Table 3).

Table 3

Interventions in eligible studies.

First author, publication year	Intervention brief description	Provider	Type of intervention	Duration of intervention
Oostdam, 2012	Warming-up such as slow cycling, individualized program of aerobic and strength exercises, cool down.	Physiotherapist	Individualised; supervised	From 15 wks of gestation to delivery
	Equipment: cycle ergometers, treadmills, cross-trainers, stationary rowing machines, free weights, accelerometer.
	Motivation component: Information on the benefits for mother and child at the start and during the intervention.
Price, 2012	Aerobic training 4 times per week, 3 times at moderate intensity as a group, consistent with exercise guidelines of the ACOG. Also, walk individually once weekly.	Researchers	Both as a group and individually; supervised	From 12–14 week of gestation to 36 week of gestation or to delivery if participants wished
	Equipment: Treadmills, elliptical trainers, stationary bicycles, weight machines, exercise balls.
	Motivation: not included
Barakat, 2013	Aerobic dance activities of 3–4 min with 1 min breaks, moderate-intensity resistance exercises lasted 25–30 min, warm-up and cool-down period both of 10–12 min duration with standards of the American College of Obstetricians and Gynaecologists.	Fitness specialist with the assistance of an obstetrician	In a group; supervised	From weeks 10 to 12 of pregnancy to the end of the third trimester (weeks 38–39)
	Equipment: Bar-bells, therabands, heart rate monitor
	Motivation: All sessions were accompanied with music, and were performed in an airy, well-lighted exercise room at the Hospital.
Nobles, 2015	30 minutes or more of moderate-intensity physical activity on most days of the week. Specific activities self-selected and including dancing, walking, and yard work.	Health educators	ND	10 wks on avarage
	Equipment: Digital pedometer.
	Motivation: Booster telephone calls and tip sheets mailed.
Seneviratne, 2015	Cycling home based moderate-intensity exercise sessions. Each exercise session included a 5-minute warm-up and cool-down period at low intensity. Frequency varyed between three and five sessions per week, and duration between 15 and 30 minutes per session, according to stage of pregnancy.	Exercise physiologist	Individualised; unsupervised	From 20 to 35 weeks of gestation
	Equipment: Magnetic stationary bicycles, heart rate monitors
	Motivation: not included
Guelfi, 2016	5-minute warm up of pedaling, 5-min periods of continuous moderate-intensity cycling alternating with 5-min periods of interval cycling, 5-minute cool down followed by light stretching.	Exercise physiologist	Individualised; supervised	14 wks
	Equipment: Upright cycle ergometer, accelerometer.
	Motivation: not included
Krohn Garnæs, 2016	Three times a week 35 minutes of moderate-intensity endurance exercise and 25 minutes of strength training. Determination of the endurance exercise at 80% of the maximum capacity, according to the Borg scale 12–15.	Physiotherapist	In a group; supervised	From 12th-18th gestational week to delivery
	Equipment: Τreadmill.
	Motivation: Motivational interview session, either individually or in a group and encouragement to compare their own weight gain with the recommended.
Wang, 2017	Exercise at the beginning of the intervention at the lower calculated limit, based on the maximum predicted heart rate for age, progressively increased with the progress of the program, at least 3 days a week.	Researchers	Supervised	27 ± 2 wks
	Equipment: Stationary bike.
	Motivation: not included
Daly, 2017	10-minute warm-up, 15–20 minutes of resistance or weights, 15–20 minutes of aerobic exercises, and a 10-minute cool-down.	Researchers	In a group; supervised	13 4/7 ± 1 2/7 wks of gestation
	Equipment: Weights.
	Motivation: Goal-setting and journaling of varied classes each day to maintain interest. Women in the intervention arm also received an invitation to a secret Facebook group to create a sense of community among participants, to share healthy lifestyle advice, and to improve compliance with the exercise intervention.

wks, weeks; ND, no data

Reporting of exercise intervention in eligible studies

Based on CERT [23], we captured the number of studies with inadequate reporting of the description of the exercise intervention (S2 Table). There was no trial that provided with adequate information for exercise reproduction. Six [2, 5, 31–33, 36] out of nine papers did not provide any information on the content of home program component. Six trials [5, 13, 31, 33–35] did not adequately report on non-exercise components. Generally, all trials provided information on the exercise components, and the necessary equipment; on the provider, and the supervision of the intervention; as well as on adherence, on potential side-effects /adverse events, and on dosage.

Effectiveness and safety of exercise during pregnancy

GDM was diagnosed by measuring fasting blood glucose, hemoglobin A1c, or by an oral glucose tolerance test (Table 4). A total of 374 (24.8%) [160 (21.4%) in intervention, and 214 (28.1%) in control group] developed GDM among the 1,508 high-risk women analysed for GDM outcome (Fig 2). When the nine trials were combined, there was no between study heterogeneity (Q 10.08, P-value 0.26). However, we could not exclude large variability (upper limit for Ι2 > 50%) in study effects due to real study differences [I2 21% (95%CI 0, 62%)]. Thus, random effects estimates would be more appropriate for data synthesis and fixed effects estimates were not presented. Women who received exercise during pregnancy were on average less likely to develop GDM compared to women who followed only the standard prenatal care (OR 0.70, 95%CI 0.52, 0.93; P-value 0.02) (Fig 2).

Table 4

Tests used for the diagnosis of GDM, and reported side effects /adverse events in the eligible RCTs.

First author, publication year	Diagnostic test for GDM	Side effects /adverse events (intervention /control)
Oostdam, 2012	FBG, HbA1c	none reported
Price, 2012	50-g 1hr OGTT	anxiety with exercise (1 / 0); history of preterm pregnancy (1 / 0); pain from leiomyomas (1 / 0)
Barakat, 2013	75-gr OGTT 2hr	premature labour (5 / 3); pregnancy-induced hypertension (5 / 4); persistent bleeding (3 / 0); molar pregnancy (0 / 3)
Nobles, 2015	50-gr 1hr OGTT	developed medical contraindication (3 / 1); miscarriage or termination (1 / 2)
Seneviratne, 2015	75-gr OGTT FBG and / or 2hr	none reported
Guelfi, 2016	FBG, 75-gr OGTT 2hr, or both	pregnancy loss (1 / 2)
Krohn Garnæs, 2016	FBG or 120-min BG	none reported
Wang, 2017	75-gr 2hr OGTT	cervical length < 25 mm (1 / 5); other*
Daly, 2017	75-gr 2hr OGTT	none reported

GDM, gestational diabetes mellitus; OR, odds ratio; CI, confidence interval; FBG, fasting blood glucose; HbA1c, hemoglobin A1c; min, minutes; BG, blood glucose; OGTT, oral glucose tolerance test; ND, no data

* other included 4 side effects / adverse events (intervention / control): ankle sprain (1 / 0); low-lying placenta (1 / 0); fetal death in utero (0 / 1); malformation (0 / 1)

Fig 2

Exercise intervention and the risk of gestational diabetes mellitus in high-risk pregnant women.

Each study is shown by an odds ratio (OR) estimate, along with ‘whiskers’ corresponding to its 95% confidence interval (95% CI). Studies are ordered according to the weight they contributed to the meta-analysis. The summary OR by random effects calculations is also shown.

The summary odds ratio showed also a significant effect when analyses were limited to studies with more than 5% of the participating women reporting a low education level (OR 0.55, 95%CI 0.40, 0.74; P-value 0.0001); studies reporting the use of a motivation component in the intervention (OR 0.69, 95%CI 0.50, 0.96; P-value 0.03); and studies that evaluated an intervention with duration more than 20 weeks (OR 0.54, 95%CI 0.40, 0.74; P-value 0.0001). However, the test of difference was significant only for the subgroup analysis based on exercise duration (studies with exercise duration more than 20 weeks vs. studies with duration up to 20 weeks; P-value = 0.02) (S3 Table). In sensitivity analyses, the summary odds ratio remained statistically significant when studies were limited to those with a low attrition bias (OR 0.70, 95%CI 0.51, 0.97; P-value 0.03). We could not exclude large variability in study effects due to real study differences for all subgroup and sensitivity analyses (S3 Table). Thus, even statistically significant effects should be interpreted with caution because the true differences in effects across studies might be due to unidentified or unexplained underlying factors. Meta-regression analyses with baseline risk, and study duration as covariates did not show a statistically significant effect on the summary OR (S4 Table).

Pregnancy-induced hypertension was the most frequently reported adverse event. Four trials [13, 31, 32, 34] reported that there was no adverse event (Table 4).

Quality of reporting, potential bias, and quality of evidence

There was good agreement between the two independent researchers [Cohen k 91.4% (95% CI 82.8%, 100%; P-value <0.001)]. Based on the overall risk, four trials were judged to raise some concerns because they failed to report specific quality domains. Specifically, two trials [2, 35] did not provide information on participants and personnel blinding, and on blinding of outcome assessors; and two studies [13, 31] did not provide information on participants and personnel blinding only. Three RCTs were judged to be at high risk of bias. One of them [36] did not provide information on participants and personnel blinding, and on blinding of outcome assessors; in addition, it reported a drop-out rate at 31.9%. The other two trials [5, 32] were unblinded for participants and personnel; one of them [5] was also unblinded for outcome assessors (Table 5).

Table 5

Quality of reporting for eligible studies.

First author, publication year	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)	Other bias
Oostdam, 2012	L	L	?	L	L	L	L
Price, 2012	L	L	?	?	H	L	L
Barakat, 2013	L	L	H	L	L	L	L
Nobles, 2015	L	L	?	?	L	L	L
Seneviratne, 2015	L	L	H	L	L	L	L
Guelfi, 2016	L	L	?	?	L	L	L
Krohn Garnæs, 2016	L	L	?	L	L	L	L
Wang, 2017	L	L	H	H	L	L	L
Daly, 2017	L	L	H	L	L	L	L

H, high risk; L, low risk;?, unclear

Based on the funnel plot assessment, there was variation in the standard error of the studies. However, small studies were reasonably closely distributed around the summary effect estimate [29] (Fig 3). Egger’s test of small study effects had a P-value of 0.31, and thus, it was not fully conclusive.

Fig 3

Funnel plot including all studies comparing an exercise intervention vs. standard prenatal care for gestational diabetes prevention among pregnant high-risk women [P-value 0.68 in the weighted regression of ln(OR) against the standard error].

Five out of the 9 studies had potential performance, detection, or attrition bias. The other 4 studies were unclear about blinding. Overall, moderate quality of evidence showed that exercise during pregnancy for the population of women with high risk for GDM may have benefit when compared to standard prenatal care in reducing the risk of GDM (S5 Table).

Discussion

Our study showed that on average an exercise intervention during pregnancy may have a beneficial effect in preventing high-risk pregnant women from developing GDM. There was no significant between study heterogeneity. However, we noticed that a large variability in study effects could not be excluded. A potential beneficial effect was also supported when analyses were limited to studies with more than 5% of the participating women reporting a low education level; studies reporting the use of a motivation component in the intervention; and studies that evaluated an intervention with duration more than 20 weeks. Subgroup and sensitivity analyses did not identify a clinical or methodological factor that may explain for the potential large variability.

Our meta-analysis supported the possibility that specific exercise programs during pregnancy may decrease the GDM incidence. Exercise programs should follow guidelines for designing complex interventions [21]. Based on CERT [23], reporting of several intervention characteristics was missing. The study [5] with a significant decrease in GDM incidence did not provide data on whether the intervention was in group or applied individually; on any motivation strategies, on the content of home exercise, and on other non-exercise components; and on whether the exercise intervention was individually tailored or not. Thus, it may not be feasible for this intervention to be reproduced in future trials. Previous studies on complex interventions also showed inadequate reporting [4, 6–10, 17, 18].

Other interventions such as diet, supplements, and medications were evaluated for GDM prevention. Results regarding these outcomes also need to be scrutinized. Some of these interventions may be important but spurious effects due to various biases may be affecting these trials as well. For general population, some meta-analyses assessing exercise interventions with or without a diet component showed also statistically significant GDM risk reduction [4, 6, 7, 9, 17]. However, other studies [8, 10, 18, 19] did not support similar results. In line with our study, a meta-analysis that evaluated exercise among overweight or obese women showed a reduction in GDM incidence [20]. Another meta-analysis [37] that evaluated the effect of different types of exercise and metformin for pregnancy outcomes in overweight and obese pregnant women, showed a reduced risk for GDM with aerobic exercise. However, our subgroup analysis limited to studies that included high-risk women based on the BMI criterion, did not show a significant effect of exercise intervention on GDM incidence. Compared to previous studies, our meta-analysis followed a more pragmatic approach for population eligibility including not only studies with pregnant women with increased BMI but also studies with pregnant women with other risk factors for GDM. Based on our subgroup analysis, future research on exercise interventions with adequate duration among pregnant women might be promising. However, this result should be interpreted cautiously since a large variability in study effects could not be excluded.

Several modifiable factors as well as non-modifiable factors may contribute to GDM. Obese women had twice the risk for GDM as compared to women with normal body weight [5]. Elevated pre-pregnancy BMI is associated with complications during pregnancy, regardless of GDM onset [1, 14, 16]. A cost-effectiveness study [38] showed that promoting healthy eating and physical activity was the preferred strategy for limiting weight gain during pregnancy. However, the exact intervention components that may lead to clinically significant risk reduction are yet to be determined. Previous studies supported that women during pregnancy showed low motivation to change their lifestyle [11]. Our subgroup analysis limited to studies that included a motivation component in the intervention also supported a significant effect of exercise intervention on GDM incidence. Therefore, exercise interventions may include a behavior change component. They may also address social determinants of health including education level to improve literacy, and access to health care services in addition to biological factors, and the right timing for women to start the intervention to prevent GDM.

Our findings may support on average a protective effect of exercise intervention during pregnancy for GDM prevention among women with GDM risk factors. However, both for the main analysis and for the subgroup and sensitivity analyses, potential large true differences in effects among studies could not be excluded. There may be additional unidentified or unexplained underlying factors that may account of the differences in effects. Future large, good quality trials recruiting pregnant women of low education level and evaluating an exercise intervention with satisfactory duration need to adequately report on the intervention characteristics to allow for evaluating potential frequency-response relationship between exercise and GDM risk reduction [12]. Additionally, they need to provide with adequate description of the exercise intervention programs for their accurate reproduction [23]. Motivation techniques for participants to complete the intervention, intensive monitoring to minimize losses to follow up that are not due to miscarriage, premature delivery, or fetal death in utero, and procedures that enhance fidelity are prerequisites for adequately implementing exercise interventions. Additional efforts to ensure blindness both of participants and researchers are imperative to support robustness of the results. In previous trials, investigators found it difficult to double-blind RCTs due to the nature of the intervention [3, 11]. Previous results on diet interventions to prevent GDM were also heterogeneous [12, 13]. Future trials assessing interventions including multiple components, i.e., diet, exercise, behavioral counseling, and social support, are needed to provide with definitive answers on their benefit and sustainability [10, 18].

Our study had several limitations. We included only studies that evaluated interventions initiated during pregnancy; therefore, our findings cannot be generalized to exercise interventions that may begin before pregnancy. However, this limited the heterogeneity of the duration of intervention among studies. We included RCTs that recruited only high-risk women; and therefore, our results cannot be generalized to general population. However, we considered as high-risk not only women who were overweight or obese but also women with other risk factors including ethnicity, medical, and family history, and sedentary lifestyle. By broadening the criteria, we tried to achieve a pragmatic approach of the population included in our work. Searching for grey literature might have identified additional studies; however, unpublished results would still have remained unknown.

Conclusion

As a conclusion, our study may support a beneficial effect of exercise interventions during pregnancy in addition to standard antenatal care for preventing GDM among high-risk women. Furthermore, a protective effect for specific population subgroups, i.e., women with low education level, and for interventions with specific characteristics, i.e., with more than 20 weeks duration, and with motivational strategies cannot be excluded. Future large, good quality studies focusing on specific women populations, and evaluating interventions with adequate duration are necessary.

Search strategy.

(DOCX)

Click here for additional data file.

Evaluation of the exercise intervention based on Consensus on Exercise Reporting Template (CERT) tool.

(DOCX)

Click here for additional data file.

Subgroup and sensitivity analyses.

(DOCX)

Click here for additional data file.

Meta-regression results for GDM OR.

(DOCX)

Click here for additional data file.

GRADE evaluation of overall evidence.

(DOCX)

Click here for additional data file.

PRISMA checklist.

(DOCX)

Click here for additional data file.

17 Jan 2022

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Carsten Bogh Juhl, PhD

Academic Editor

PLOS ONE

Journal requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf”.

Additional Editor Comments:

Dear author – thank you for your systematic review and meta-analysis on an important health problem. However some issues remain as pointed out by the reviewers. Further I have some additional issues that may need to be addressed before publication.

In the literature search was the MeSH terms used or was the search just performed as text words. The use of the filter – randomized controlled trial – may be less sensitive than the Cochrane highly sensitive filter for identifying RCT (these issues need to be addressed as limitation)

The study was not protocolized in either open science framework (www.osf) nor in PROSPERO. This is a severe limitation and need to be addressed too.

The study was reported according to the PRISMA guidelines – using the PRISMA flowchart may improve the readability and the quality of the flowchart. Further the reporting in the flowchart is inconsistent – 780 hits was retrieved and 72 was duplicates – but only 474 was screened – and 639 were excluded based on title and abstract. Further some difference between text and flowchart are seen (73 or 72 duplicates).

The author used Odds Ratio (OR) for pooling the results – even though OR showed more extreme results compared to Relative Risk (RR) and all included studies are RCT making an meta-analysis on RR possible. Presenting the results in RR may easy the interpretability of the results for clinician. Further the plots are quite blurry (both the forest plot and the flowchart).

The author stated that they has especial focus on explaining the heterogeneity - however why is the only analysis for investigating heterogeneity Eggers test for small study bias – what about the amount of exercise – the duration of the exercise intervention – the risk of bias – et cetera. These analysis can be performed comparing subgroups or investigated in a meta-analysis.

Finally the GRADE approach should be used for evaluating the overall evidence (based on study limitation, inconsistency, indirectness, impression and publication bias)

Best Associate Professor Carsten Juhl

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: No

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Εffectiveness of exercise intervention during pregnancy on high-risk women for gestational diabetes mellitus prevention: a meta-analysis of published RCTs

In this meta-analysis of published RCTs on exercise in pregnant women at risk of GDM, the authors find a significant benefit of exercise in preventing GDM but prudently infer cation to their results due to huge heterogeneity of the ten included RCT studies. The meta-analysis contributes to the existing body of literature on the topic.

The strengths of the study include a strong adherence to reporting guidelines such as PRISMA and standardized methods such as CERT and tools proposed by the Cochrane Collaboration. The method section is rigorous and well written. The authors have a pragmatic approach to the risk factor for GDM in the included population which is well elucidated. The weaknesses of the included studies are clarified and well put in the context of the meta-analysis.

The shortcoming of the manuscript is the difficulty of addressing the impact of the findings and viewing the results in the context of similar studies. What is the impact of this meta-analysis compared to the other meta-analysis on the topic?

There are also a few more questions that require addressing:

Abstract page 3: The last four lines of the Results section are repeated in the conclusion. Consider omitting the four lines in the results section to include more results from the meta-analysis.

The first line of the conclusion should address the aim of the study.

Introduction page 4: The authors mention several risk factors for GDM but in the research question it is not clear how they define “high-risk pregnant women”. Do the women have 1, 2, 3 4 or all of the mentioned risk factors to be at high risk. The definition of “high-risk pregnant women” should be elucidated already in the introduction?

Methods section (Eligibility Criteria) page 5: Ethnicity as a risk factor could warrant elaboration since it is not as intuitive as the other risk factors.

Discussion section page 15-16: To strengthen the discussion section, it would be beneficial with a more direct comparison of your study to [ref 18] and [ref 19].

[ref 18]: This meta-analysis found no effect of physical exercise on GDM. Why is their conclusion different to yours? Differences in eligibility criteria for included studies or differences in the characteristics of participating women in the eligible studies?

[ref 19]: This meta-analysis also finds a beneficial effect of exercise but with fewer reservations about their result. Is it only because of prenatal exercise or small sample sizes in included studies? Be kind to elaborate – what does the author's scepticism consist of?

Discussion section page 17: The authors have to further elaborate on how this study adds to what is already known on exercise to prevent GDM apart from identifying the presence of publication bias. However, the authors succeed in suggesting improvements in future RCT such as an adequate description of the exercise intervention and motivational techniques for participants to complete the intervention.

Reviewer #2: This article addresses the preventing effect of exercise among women with high risk of gestational diabetes mellitus during pregnancy. Which is an important topic.

There are some areas in this article, that needs to be addressed/clarified. In general, the writing quality needs to be improved to a higher and more professional level.

The paper is with sufficient details.

Abstract:

Overall the abstract is relevant and capture the articles.

The phrase “between-study heterogeneity was estimated” is used multiple times during the article.

Ex. page 3: “There was significant between study heterogeneity [Q 24.45, P-value 0.004; I2 = 63% (95%CI 27%, 81%)]”.

It seems like you have mixed two expressions, heterogeneity is between study variance above the expected (chance), so you have to decide to write between study variance or heterogeneity.

Besides that, one should separate the values, so one cite the Q and I value for themself. because I is the proportion of the total variation due to inconsistency, and Q is saying if the variation is random (If Q is significant (p < 0.05) => heterogeneity).

Introduction:

In the introduction it would be relevant to mention education levels impact on GDM, since it is mentioned in table 2 later.

On page 4 it is stated that “We performed meta-analysis with special emphasis on issues of potential biases and sources of heterogeneity between studies.” Here it would be good to analyze what the heterogeneity is, and not only if there is publication bias.

Methods:

Some parts of the Methods section are not adequately explained

In the introduction one writes “Identified risk factors for GDM include obesity [1,4-6,9-13], sedentary lifestyle [10], unbalanced diet [3,10], ethnicity [2,4,10,14] and family history [4,10] “and then in the Eligibility Criteria you elaborate, but it is highly recommended to elaborate on ethnicity as well.

Besides that, sedentary lifestyle and unbalanced diet is mentioned as a risk factor, but not used as an eligibility Criteria. What are your thoughts about that specific? A small comment about this would be good to include.

In the data extraction on page 6 you mention that you extract level of education, but its not used it for anything, please think about why you extract it.

Under the quality assessment of the studies its written that: “We used the risk of bias tool proposed by the Cochrane Collaboration for quality assessment of eligible RCTs “. But it’s not used for anything. Its only too state the findings, but not what it means for the interpretation of the study as a whole.

It’s not mentioned how many authors asses the quality of the studies, this is first mentioned during the presentation of the results, consider to mention this in method section instead of in the results. Besides that, you should mention what outcome you asses.

Under statistical Analysis on page 6 you write that “Heterogeneity was evaluated with Cochran’s Q statistic (statistically significant for P < 0.10)” and later you write “The level of significance was set at P < 0.05”. Hence, it would be good to clarify what significant level that have been used.

Results:

In general, avoid re-iterating all the results from the table by text, instead just highlight the principal findings.

Data presentation, for a more manageable data presentation consider simplified the tables and layout.

In table 2 (Characteristics of participating women in the eligible studies) you include low education level – but have not mentioned it earlier on – and its not used after – so the information doesn’t seem relevant.

In the quality assessment of the studies: You assessed the risk of bias, but not the the overall risk of bias, which is used to know the validity of the results, so it would be good to present it.

Under the effectiveness and safety of exercise during pregnancy: page 11, its mentioned between study heterogeneity again. See earlier comment about heterogeneity

You present odds ratio, Participants analyzed intervention /control and GDM events, and n (%) intervention /control, in table 4 and figure 2, There is no need to present that twice, only present it once.

On page 14, in table 5 (other bias), you have 7 studies that get high risk of bias, but you don’t state why.

Discussion

There is some overlapping between Introduction and Discussion sections.

The reason for heterogeneity is not discuss, even though you earlier have stated this “We performed meta-analysis with special emphasis on issues of potential biases and sources of heterogeneity between studies”. Therefore, it would make sense to analyze and elaborate what causes the heterogeneity.

You write: “A subsequent 16 meta-analysis [18] did not show any benefit of exercise, metformin, vitamin D, and probiotics as compared to placebo or no treatment among high-risk women. These conflicting results indicated that there should be some skepticism about trials on GDM preventive interventions.”

The article: Pascual-Morena C, Cavero-Redondo I, Álvarez-Bueno C, et al. Exercise versus Metformin to Improve Pregnancy Outcomes among Overweight Pregnant Women: A Systematic Review and Network Meta-Analysis. J Clin Med. 2021;10(16):3490. Published 2021 Aug 7. doi:10.3390/jcm10163490. Find that aerobic exercise showed an effect on GDM– which support your finding. – how does that effect the believe in you finding?

You write that future studies should: intensive monitoring to minimize losses to follow up

- but a lot of the events that caused loss is due to change in pregnancy like miscarried/ premature delivery or fetal death in utero, and those events you can’t change.

In the result section you used a lot of space comment on the “Characteristics of the included studies”, but you do not comment on what this means for the results in the article.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

25 Mar 2022

Responses to Editor and Reviewers

Editor’s comments

1) In the literature search was the MeSH terms used or was the search just performed as text words. The use of the filter – randomized controlled trial – may be less sensitive than the Cochrane highly sensitive filter for identifying RCT (these issues need to be addressed as limitation).

REPLY: We thank the Editor for this comment. We performed again the search in the three databases according to the Editor’s suggestion. We rephrased to “For Pubmed, we used a search strategy including keywords related to exercise, physical activity, and GDM combined with the Cochrane Collaboration search algorithm for RCTs. We conducted a systematic search on Scopus using the same keywords after excluding articles registered in Pubmed. Finally, we searched CENTRAL including the same keywords related to exercise, physical activity, and GDM. Search algorithms were described in detail in Table S1.” [In Material and methods, Search strategy]

2) The study was not protocolized in either open science framework (www.osf) nor in PROSPERO. This is a severe limitation and need to be addressed too.

REPLY: As suggested, we protocolized the study in the open science framework (www.osf). Specifically, we added that “Our study was pre-registered in the Open Science Framework (OSF) (Registration DOI 10.17605/OSF.IO/23NJS, https://archive.org/details/osf-registrations-23njs-v1).” [In Material and methods, first paragraph]

3) The study was reported according to the PRISMA guidelines – using the PRISMA flowchart may improve the readability and the quality of the flowchart. Further the reporting in the flowchart is inconsistent – 780 hits was retrieved and 72 was duplicates – but only 474 was screened – and 639 were excluded based on title and abstract. Further some difference between text and flowchart are seen (73 or 72 duplicates).

REPLY: We thank the Editor for this comment. After performing the new search strategy, we realized during the screening that several of the RCTs that were included as eligible in the manuscript we initially submitted, evaluated an intervention that included a diet component in addition to exercise. This was in contrast with our initial inclusion criteria. Since those trials were erroneously included in the first manuscript, after our revised search, we excluded them. We have now presented the correct numbers in the text “Our search yielded 1267 items (458 in PubMed, 235 in Scopus, and 574 in CENTRAL). We excluded 216 as duplicated. Out of the 1051 remaining items, we excluded 1019 as non-relevant based on the title, or abstract. Thus, we retrieved 32 papers in full text. Out of the 32 articles, we excluded 25; one paper reported a pilot study; five studies did not include an eligible population; 7 studies included a non-eligible intervention; and 12 trials did not report the onset of GDM as an outcome. Finally, we included 7 published RCTs as eligible for our study (Figure 1)”. In addition, we revised the flow chart (Figure 1).

4) The author used Odds Ratio (OR) for pooling the results – even though OR showed more extreme results compared to Relative Risk (RR) and all included studies are RCT making a meta-analysis on RR possible. Presenting the results in RR may easy the interpretability of the results for clinician.

REPLY: We agree with the Editor that RR may have easy interpretability. However, odds ratios have favorable mathematical properties; and therefore, we chose to complete our analyses with OR. In addition, the assumed control risk (ACR 0.30) cannot be considered high. Thus, the two estimates (OR and RR) coincide (OR = RR = 0.70).

5) Further the plots are quite blurry (both the forest plot and the flowchart).

REPLY: As suggested, we revised both the forest plot (Figure 2) and the flowchart (Figure 1) to improve image resolution.

6) The author stated that they have especial focus on explaining the heterogeneity - however why is the only analysis for investigating heterogeneity Eggers test for small study bias – what about the amount of exercise – the duration of the exercise intervention – the risk of bias – et cetera. These analysis can be performed comparing subgroups or investigated in a meta-analysis.

REPLY: As suggested, we added that “We performed separate analyses limited to studies where increased BMI was included in as a risk factor for GDM, and studies that did not consider BMI; studies where the percentage of participating women with low level education was more than 5%; studies that evaluated an intervention delivered individually, and studies that evaluated an intervention delivered in a group; trials that included a motivation component in the intervention, and trials that did not include motivation; studies with an intervention duration more than 20 weeks, and studies with an intervention duration up to 20 weeks. We also performed meta-regression analyses on GDM OR. The effect of baseline risk, and study duration were included individually as covariates in the meta-regressions. For each meta-regression, the slope coefficient with the standard error (SE), the permutation-based P value (as suggested by Higgins and Thompson [27] and the tau2 were reported” [In Material and methods, Statistical analysis]

We also added “The summary odds ratio did not change much when sensitivity analyses were limited to studies fulfilling specific criteria including the use of increased BMI as a risk factor for GDM; the use of a motivation component in the intervention; and the type of intervention delivery (individually, or in a group) (Table S3). However, the summary odds ratio suggested a protective effect for studies with more than 5% of the participating women reporting a low education level (OR 0.49, 95%CI 0.43, 0.73; P-value 0.0006); and for studies that evaluated an intervention with duration more than 20 weeks (OR 0.47, 95%CI 0.31, 0.73; P-value 0.0007). Τhere was no between study heterogeneity for all sensitivity analyses except for one (Table S3). However, we could not exclude large variability in study effects due to real study differences for all sensitivity analysis. (Table S3) Thus, even statistically significant effects should be interpreted with caution because the true differences in effects across studies might be due to unidentified or unexplained underlying factors. Meta-regression analyses with baseline risk, and study duration as covariates did not show a statistically significant effect on the summary OR (Table S4)” [In Results, Effectiveness and safety of exercise during pregnancy]

Finally, we added two supplementary tables (Table S3 for subgroup and sensitivity analyses; and Table S4 for meta-regression analyses).

7) Finally, the GRADE approach should be used for evaluating the overall evidence (based on study limitation, inconsistency, indirectness, impression and publication bias)

REPLY: We thank the Editor for this comment. As suggested, we added “In addition, we used the Grading of Recommendations, Assessment, Development and Evaluation tool (GRADE) for rating the overall evidence [24] (GRADEpro, Version 3.6.1 McMaster University, 2011)”. [In Material and methods, Quality assessment of the studies and rating of overall evidence]

We also added “Overall, moderate quality of evidence showed that exercise during pregnancy in high-risk women may not have benefit when compared to standard prenatal care in reducing the risk of GDM. The level of evidence for RCTs was downgraded due to the unreported or lack of blinding in participants, personnel, and outcome assessors in most of the studies and because the optimal information size was not met.” [In Results, Quality of reporting, potential bias, and quality of evidence] We also added Table S5.

Reviewer 1

1) The shortcoming of the manuscript is the difficulty of addressing the impact of the findings and viewing the results in the context of similar studies. What is the impact of this meta-analysis compared to the other meta-analysis on the topic?

REPLY: We thank the reviewer for this comment. We added that “However, to our knowledge, there was no systematic approach to evaluate exercise as a single intervention during pregnancy on GDM prevention among high-risk women with any of the risk factors for GDM, and who already received standard prenatal care.” [In Introduction, third paragraph, last sentence]

2) Abstract page 3: The last four lines of the Results section are repeated in the conclusion. Consider omitting the four lines in the results section to include more results from the meta-analysis. The first line of the conclusion should address the aim of the study.

REPLY: As suggested, we omitted the last four lines in the Results section in Abstract. We also rephrased our Conclusion to “Our study did not support a protective effect of exercise intervention during pregnancy for high-risk women to prevent GDM.” [In Abstract, Conclusions]

3) Introduction page 4: The authors mention several risk factors for GDM but in the research question it is not clear how they define “high-risk pregnant women”. Do the women have 1, 2, 3 4 or all of the mentioned risk factors to be at high risk. The definition of “high-risk pregnant women” should be elucidated already in the introduction?

REPLY: As suggested, we clarified in the research question (last paragraph in Introduction) that “We included RCTs on high-risk pregnant women with one or multiple risk factors…”.

4) Methods section (Eligibility Criteria) page 5: Ethnicity as a risk factor could warrant elaboration since it is not as intuitive as the other risk factors.

REPLY: We thank the reviewer for this comment. As suggested, we clarified “…non-white ethnicities…”. [In Methods, Eligibility criteria]

5) Discussion section page 15-16: To strengthen the discussion section, it would be beneficial with a more direct comparison of your study to [ref 18] and [ref 19].

[ref 18]: This meta-analysis found no effect of physical exercise on GDM. Why is their conclusion different to yours? Differences in eligibility criteria for included studies or differences in the characteristics of participating women in the eligible studies?

REPLY: We clarified that “A meta-analysis that evaluated exercise initiated before pregnancy among overweight or obese women showed a reduction in GDM incidence [19]. However, our study focused on RCTs that evaluated an exercise intervention during pregnancy, which might shed light on whether interventions during that period are worth being considered. Based on our subgroup analysis, future research on exercise interventions with adequate duration among pregnant women might be promising. … In line with our study, a subsequent meta-analysis [18] did not show any benefit of exercise, metformin, vitamin D, and probiotics as compared to placebo or no treatment among obese pregnant women. However, our meta-analysis followed a more pragmatic approach for population eligibility including not only studies with pregnant women with increased BMI but also studies with pregnant women with other risk factors for GDM.” [In Discussion, paragraph 4]

6) [ref 19]: This meta-analysis also finds a beneficial effect of exercise but with fewer reservations about their result. Is it only because of prenatal exercise or small sample sizes in included studies? Be kind to elaborate – what does the author's scepticism consist of?

REPLY: As suggested, we clarified “Based on our subgroup analysis, future research on exercise interventions with adequate duration among pregnant women might be promising. However, this result should be interpreted cautiously since a large variability in study effects could not be excluded.” [In Discussion, paragraph 4]

7) Discussion section page 17: The authors have to further elaborate on how this study adds to what is already known on exercise to prevent GDM apart from identifying the presence of publication bias. However, the authors succeed in suggesting improvements in future RCT such as an adequate description of the exercise intervention and motivational techniques for participants to complete the intervention.

REPLY: As suggested, we rephrased “Our findings could not support on average a protective effect of exercise intervention during pregnancy for GDM prevention. Based on our sensitivity analyses, a beneficial effect cannot be excluded among women with low education level, and for exercise interventions with a more than 20-week duration. However, even for these two analyses, potential large true differences in effects among studies could not be excluded. There may be additional unidentified or unexplained underlying factors that may account of the differences in effects. Future large, good quality trials recruiting pregnant women of low education level and evaluating an exercise intervention with satisfactory duration need to adequately report on the intervention characteristics to allow for evaluating potential frequency-response relationship between exercise and GDM risk reduction [12].” [In Discussion, paragraph 6]

Reviewer 2

Abstract:

1) The phrase “between-study heterogeneity was estimated” is used multiple times during the article.

REPLY: As suggested we explained that “Between study heterogeneity (Cochran’s Q statistic) and the extent of study effects variability [I2 with 95% confidence interval (CI)] were estimated.” [In Abstract, Methods section]

We also clarified the phrase “…between study heterogeneity…” throughout the whole manuscript.

2) Ex. page 3: “There was significant between study heterogeneity [Q 24.45, P-value 0.004; I2 = 63% (95%CI 27%, 81%)]”.

It seems like you have mixed two expressions, heterogeneity is between study variance above the expected (chance), so you have to decide to write between study variance or heterogeneity. Besides that, one should separate the values, so one cite the Q and I value for themself. because I is the proportion of the total variation due to inconsistency, and Q is saying if the variation is random (If Q is significant (p < 0.05) => heterogeneity).

REPLY: As suggested, we rephrased to “Τhere was no between study heterogeneity (Q 0.90, P-value 0.64 for low education level; and Q 0.82, P-value 0.84 for duration more that 20 weeks). However, we could not exclude large variability in study effects (I2 0; 95%CI 0, 90%; and I2 0; 95%CI 0, 85% respectively)”. [In Abstract, Results section]

Introduction:

3) In the introduction it would be relevant to mention education levels impact on GDM, since it is mentioned in table 2 later.

Reply: As suggested, we added “Identified risk factors for GDM include…socioeconomic factors including low education level [14]…” [In Introduction, 2nd paragraph]

4) On page 4 it is stated that “We performed meta-analysis with special emphasis on issues of potential biases and sources of heterogeneity between studies.” Here it would be good to analyze what the heterogeneity is, and not only if there is publication bias.

REPLY: As suggested, we rephrased to “We performed meta-analysis with special emphasis on … and sources of study heterogeneity including both clinical and methodological factors that may account for potential variability in study effects.” [In Introduction, last paragraph]

Methods:

5) Some parts of the Methods section are not adequately explained

In the introduction one writes “Identified risk factors for GDM include obesity [1,4-6,9-13], sedentary lifestyle [10], unbalanced diet [3,10], ethnicity [2,4,10,14] and family history [4,10] “and then in the Eligibility Criteria you elaborate, but it is highly recommended to elaborate on ethnicity as well. Besides that, sedentary lifestyle and unbalanced diet is mentioned as a risk factor, but not used as an eligibility Criteria. What are your thoughts about that specific? A small comment about this would be good to include.

REPLY: We thank the reviewer for this comment. As suggested, we clarified “…non-white ethnicities…”. We also rephrased to “Factors that increased pregnant women’s risk included at least one of the following: increased BMI [1,4-6,9-13], sedentary lifestyle [10], family history [4,10,21], previous macrosomia [21], unbalanced diet [3,10], previous GDM [21], non-white ethnicities [2,4,10,14,21] and age > 25 years [21].” [In Methods, Eligibility criteria]

6) In the data extraction on page 6 you mention that you extract level of education, but it’s not used it for anything, please think about why you extract it.

REPLY: We thank the reviewer for this comment. We added that “We also performed separate analyses limited to … studies where the percentage of participating women with low education level was more than 5%.” [In Methods, Statistical analysis]

7) Under the quality assessment of the studies its written that: “We used the risk of bias tool proposed by the Cochrane Collaboration for quality assessment of eligible RCTs “. But it’s not used for anything. Its only too state the findings, but not what it means for the interpretation of the study as a whole.

REPLY: We thank the reviewer for this comment. As suggested, we added that “We performed separate analyses for studies with low detection bias (studies reporting blinding of outcome assessors); and for studies with low attrition bias (studies with less than 20% of participants lost in follow-up).” [In Methods, Statistical analysis]

8) It’s not mentioned how many authors asses the quality of the studies, this is first mentioned during the presentation of the results, consider to mention this in method section instead of in the results. Besides that, you should mention what outcome you asses.

REPLY: As suggested, we added that “Two independent researchers (GIT and KP) extracted the data on quality assessment. Discrepancies were resolved with consensus, and the participation of a third arbitrator (AT) where necessary.” [In Methods, Quality assessment of the studies]

We also added that “Finally, we recorded the number of GDM events as the outcome, separately in the experimental and the control arm. We also captured information on the method used in each study for the diagnosis of GDM.” [In Methods, last paragraph of Data extraction]

9) Under statistical Analysis on page 6 you write that “Heterogeneity was evaluated with Cochran’s Q statistic (statistically significant for P < 0.10)” and later you write “The level of significance was set at P < 0.05”. Hence, it would be good to clarify what significant level that have been used.

REPLY: According to the reviewer, we clarified the “The level of significance for all analyses, except for Cochran’s Q statistic, was set at P-value < 0.05”. [In Methods, Statistical analysis]

Results:

10) In general, avoid re-iterating all the results from the table by text, instead just highlight the principal findings.

REPLY: As suggested, we rephrased our text in Results so that it highlights the principal findings.

11) Data presentation, for a more manageable data presentation consider simplified the tables and layout.

REPLY: As suggested, we tried to simplify the tables and layout to facilitate data presentation.

12) Characteristics of eligible studies:

In table 2 (Characteristics of participating women in the eligible studies) you include low education level – but have not mentioned it earlier on – and it’s not used after – so the information doesn’t seem relevant.

REPLY: We added “We performed separate analyses limited to studies … where the percentage of participating women with low level education was more than 5%...”. [In Methods, Statistical analysis

We also added “…the summary odds ratio suggested a protective effect for studies with more than 5% of the participating women reporting a low education level (OR 0.49, 95%CI 0.43, 0.73 P-value 0.0006);” [In Results, Effectiveness and safety of exercise during pregnancy]

13) In the quality assessment of the studies: You assessed the risk of bias, but not the the overall risk of bias, which is used to know the validity of the results, so it would be good to present it.

REPLY: As suggested, we rephrased “Based on the overall risk, four trials were judged to raise some concerns because they failed to report specific quality domains. Specifically, three trials [2,13,32] did not provide information on participants and personnel blinding, and on blinding of outcome assessors; and one study [30] did not provide information on participants and personnel blinding only. Three RCTs were judged to be at high risk of bias. One of them [33] did not provide information on participants and personnel blinding, and on blinding of outcome assessors; in addition, it reported a drop-out rate at 31.9%. The other two trials [5,31] were unblinded for participants and personnel; one of them [5] was also unblinded for outcome assessors. (Table 5).” [In Results, Quality of reporting and potential bias]

14) Under the effectiveness and safety of exercise during pregnancy: page 11, its mentioned between study heterogeneity again. See earlier comment about heterogeneity.

REPLY: As suggested, we rephrased to “When the seven trials were combined, there was no between study heterogeneity (Q 8.34, P-value 0.21). However, we could not exclude large variability (upper limit for Ι2 > 50%) in study effects due to real study differences [I2 28% (95%CI 0%, 69%)]”. [In Results, Effectiveness and safety of exercise during pregnancy]

15) You present odds ratio, Participants analyzed intervention /control and GDM events, and n (%) intervention /control, in table 4 and figure 2, There is no need to present that twice, only present it once.

REPLY: As suggested, we deleted the columns “Participants analyzed, intervention /control”, “GDM events, intervention /control”, and “Odds ratio (95% CI)” in Table 4.

16) On page 14, in table 5 (other bias), you have 7 studies that get high risk of bias, but you don’t state why.

REPLY: We thank the reviewer for this comment. After assessing the quality of the eligible studies, we concluded that indeed there was no high risk for the “Other bias” domain according to the Cochrane guidance on how to address the specific domain. In addition, we have changed the indication to “L” in the column “Other bias” in Table 5.

Discussion

17) There is some overlapping between Introduction and Discussion sections.

REPLY: As suggested, we rephrased the Discussion section to avoid overlapping with Introduction.

18) The reason for heterogeneity is not discuss, even though you earlier have stated this “We performed meta-analysis with special emphasis on issues of potential biases and sources of heterogeneity between studies”. Therefore, it would make sense to analyze and elaborate what causes the heterogeneity.

REPLY: We thank the reviewer for this comment. We added “Notably all trials except for one suggested no benefit, and a small study effect was unlikely. However, we noticed that a large variability in study effects could not be excluded. Several subsequent subgroup and sensitivity analyses did not identify a clinical or methodological factor that may explain for a potential large variability.” [In Discussion, paragraph 1]

19) You write: “A subsequent 16 meta-analysis [18] did not show any benefit of exercise, metformin, vitamin D, and probiotics as compared to placebo or no treatment among high-risk women. These conflicting results indicated that there should be some skepticism about trials on GDM preventive interventions.”

REPLY: We thank the reviewer for this comment. We rephrased to “In line with our study, a subsequent meta-analysis [18] did not show any benefit of exercise, metformin, vitamin D, and probiotics as compared to placebo or no treatment among obese pregnant women. However, our meta-analysis followed a more pragmatic approach for population eligibility including not only studies with pregnant women with increased BMI but also studies with pregnant women with other risk factors for GDM.” [In Discussion, paragraph 4]

20) The article: Pascual-Morena C, Cavero-Redondo I, Álvarez-Bueno C, et al. Exercise versus Metformin to Improve Pregnancy Outcomes among Overweight Pregnant Women: A Systematic Review and Network Meta-Analysis. J Clin Med. 2021;10(16):3490. Published 2021 Aug 7. doi:10.3390/jcm10163490. Find that aerobic exercise showed an effect on GDM– which support your finding. – how does that effect the believe in you finding?

REPLY: We thank the reviewer for this comment. We added “Another meta-analysis [34] that evaluated the effect of different types of exercise and metformin for pregnancy outcomes in overweight and obese pregnant women, showed a reduced risk for GDM with aerobic exercise. However, our meta-analysis followed a more pragmatic approach for population eligibility including not only studies with pregnant women with increased BMI but also studies with pregnant women with other risk factors for GDM.” [In Discussion, paragraph 4]

21) You write that future studies should: intensive monitoring to minimize losses to follow up-but a lot of the events that caused loss is due to change in pregnancy like miscarried/ premature delivery or fetal death in utero, and those events you can’t change.

REPLY: To clarify, we rephrased to “…intensive monitoring to minimize losses to follow up that are not due to miscarriage, premature delivery, or fetal death in utero…”. [In Discussion, paragraph 6]

22) In the result section you used a lot of space comment on the “Characteristics of the included studies”, but you do not comment on what this means for the results in the article.

REPLY: However, another explanation might be that exercise may be effective in specific settings, for specific population subgroups, or when specific intervention characteristics exist. Our study did not show that baseline risk for GDM, using increased BMI as a risk factor or a small percentage of women with low education level in the population might change the effect. Furthermore, including a motivation component in the intervention, the mode of intervention delivery or an intervention duration up to 20 weeks did not change the summary estimate. Based on our subgroup analyses however, a protective effect may be suggested for exercise among women with low education level, and when the intervention duration is more than 20 weeks.” [In Discussion, paragraph 2]

Submitted filename: Response to reviewers.docx

Click here for additional data file.

29 Apr 2022

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Carsten Bogh Juhl, PhD

Academic Editor

PLOS ONE

Additional Editor Comments (if provided):

Even though the authors have addressed some of the comments from the editor and the reviewer this manuscript is still insufficient for publication. The author has added the following to the abstract – however the sentence does not really make sense - However, we could not exclude large variability in study effects (I2 0; 95%CI 0, 90%; and I2 0; 95%CI 0, 85% respectively). The author stated that the study was preregistered even though the registration was performed after the first submission.

The quality of the forest plot, funnel plot and the flowchart is still too low for printing. The figures are not sufficiently self-explanatory – most of the figures need more explanation. Figure 1 presenting the search is still insufficient (does not define Mesh and text words for searching in Medline and Central – even though the author stated that the search was updated – but the one added in table one is limited up-to July 2020 and in Scopus studies from 2021 and 2022 is deleted. Table 3 are showing a subgroup analysis – however presenting the Q, I-square and p-value for the subgroup – however the test of difference may be the one of clinical interest. Table 4 is not clear and need more explanation. Table 5, the SOF table need explanation for what is study population and what does moderate means – the evidence is judged moderate even though the confidence interval is very broad and overlapping 1 (indicating no significant effect).

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: (No Response)

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: (No Response)

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: (No Response)

Reviewer #2: N/A

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: (No Response)

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: (No Response)

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: (No Response)

Reviewer #2: The protocol ismade after the first submission, which is a big limitation. The search is very deficient, and lacks Mesh terms and many relevant keywords, resulting in missing relevant articles (Unless table 1, isn't updated since last submission)

The following 3 articles would be relevant to include, and meet your inclusion criteria:

1. Barakat R, Pelaez M, Lopez C, Lucia A, Ruiz JR. Exercise during pregnancy and gestational diabetes-related adverse effects: a randomized controlled trial. Br J Sports Med. 2013 Jul; 47 (10): 630-6. doi: 10.1136 / bjsports-2012-091788. Epub 2013 Jan 30. PMID: 23365418.

(Inclusion criteria included: being sedentary, with match yours)

2. Seneviratne SN, Jiang Y, Derraik J, McCowan L, Parry GK, Biggs JB, Craigie S, Gusso S, Peres G, Rodrigues RO, Ekeroma A, Cutfield WS, Hofman PL. Effects of antenatal exercise in overweight and obese pregnant women on maternal and perinatal outcomes: a randomized controlled trial. BJOG. 2016 Mar; 123 (4): 588-97. doi: 10.1111 / 1471-0528.13738. Epub 2015 Nov 6. PMID: 26542419.

(Participants were women aged 18–40 years with a body

mass index (BMI) ≥25 kg / m2,)

3. From Oliveria Melo AS, Silva JL, Tavares JS, Barros VO, Leite DF, Amorim MM. Effect of a physical exercise program during pregnancy on uteroplacental and fetal blood flow and fetal growth: a randomized controlled trial. Obstet Gynecol. 2012 Aug; 120 (2 Pt 1): 302-10. doi: 10.1097 / AOG.0b013e31825de592. PMID: 22825089.

(The inclusion criteria consisted of: healthy pregnant women who were sedentary at admission to the

study)

Many of the previous fixes have been improved, but the search should have been significantly improved.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Jørgen Guldberg-Møller

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

12 Jun 2022

Responses to Editor and Reviewers

Editor’s comments

1) The author has added the following to the abstract – however the sentence does not really make sense - However, we could not exclude large variability in study effects (I2 0; 95%CI 0, 90%; and I2 0; 95%CI 0, 85% respectively).

REPLY: We thank the editor for this comment. We rephrased to “However, we could not exclude large variability in study effects because the upper limit of I² confidence interval was higher than 50% for all the analyses.”

2) The author stated that the study was preregistered even though the registration was performed after the first submission.

REPLY: We rephrased to “registered”. (Page 6; Paragraph 1)

3) The quality of the forest plot, funnel plot and the flowchart is still too low for printing.

REPLY: As suggested, we improved the quality of the forest plot, funnel plot and the flowchart.

4) The figures are not sufficiently self-explanatory – most of the figures need more explanation.

REPLY: We thank the Editor for this comment. We have now rephrased the Legends for the three figures to sufficiently explain their content.

5) Figure 1 presenting the search is still insufficient (does not define Mesh and text words for searching in Medline and Central – even though the author stated that the search was updated – but the one added in table one is limited up-to July 2020 and in Scopus studies from 2021 and 2022 is deleted.

REPLY: As suggested, we updated our search up to May 2022. Table S1 includes the exact MeSH terms and keywords, we used. We have also updated Figure 1. The number of articles in the Flowchart (Figure 1) correspond to the numbers we found in our last search (May 2022).

6) Table 3 are showing a subgroup analysis – however presenting the Q, I-square and p-value for the subgroup – however the test of difference may be the one of clinical interest.

REPLY: As suggested, we added the P-value for the test of difference for all subgroup analyses in the revised Table S3.

7) Table 4 is not clear and need more explanation.

REPLY: As suggested, we clarified Table 4, and provided additional explanation “GDM was diagnosed by measuring fasting blood glucose, hemoglobin A1c, or by an oral glucose tolerance test (Table 4).” (Page 14; paragraph 2)

8) Table 5, the SOF table need explanation for what is study population and what does moderate means – the evidence is judged moderate even though the confidence interval is very broad and overlapping 1 (indicating no significant effect).

REPLY: As suggested we clarified “Five out of the 9 studies had potential performance, detection, or attrition bias. The other 4 studies were unclear about blinding. Overall, moderate quality of evidence showed that exercise during pregnancy for the population of women with high risk for GDM may have benefit when compared to standard prenatal care in reducing the risk of GDM. (Table S5)” (Page 17, Paragraph 1)

Reviewer 1

All comments have been addressed.

REPLY: We thank the Reviewer for this comment.

Reviewer 2

1) The protocol is made after the first submission, which is a big limitation. The search is very deficient, and lacks Mesh terms and many relevant keywords, resulting in missing relevant articles (Unless table 1, isn't updated since last submission)

REPLY: We thank the Reviewer for this issue. As suggested, we have updated our initial search strategy including additional MeSH terms and keywords and expanding our search (from inception) to May 2022. We have updated Table S1 to include the exact search strategy. We have also updated all Tables, and Figures in the Results so that they include the additional studies. The text has also been updated as appropriate.

2) The following 3 articles would be relevant to include, and meet your inclusion criteria:

1. Barakat R, Pelaez M, Lopez C, Lucia A, Ruiz JR. Exercise during pregnancy and gestational diabetes-related adverse effects: a randomized controlled trial. Br J Sports Med. 2013 Jul; 47 (10): 630-6. doi: 10.1136 / bjsports-2012-091788. Epub 2013 Jan 30. PMID: 23365418.

(Inclusion criteria included: being sedentary, with match yours)

2. Seneviratne SN, Jiang Y, Derraik J, McCowan L, Parry GK, Biggs JB, Craigie S, Gusso S, Peres G, Rodrigues RO, Ekeroma A, Cutfield WS, Hofman PL. Effects of antenatal exercise in overweight and obese pregnant women on maternal and perinatal outcomes: a randomized controlled trial. BJOG. 2016 Mar; 123 (4): 588-97. doi: 10.1111 / 1471-0528.13738. Epub 2015 Nov 6. PMID: 26542419.

(Participants were women aged 18–40 years with a body mass index (BMI) ≥25 kg / m2,)

3. From Oliveria Melo AS, Silva JL, Tavares JS, Barros VO, Leite DF, Amorim MM. Effect of a physical exercise program during pregnancy on uteroplacental and fetal blood flow and fetal growth: a randomized controlled trial. Obstet Gynecol. 2012 Aug; 120 (2 Pt 1): 302-10. doi: 10.1097 / AOG.0b013e31825de592. PMID: 22825089.

(The inclusion criteria consisted of: healthy pregnant women who were sedentary at admission to the

study)

REPLY: Indeed, we have identified the three papers the Reviewer has indicated. As suggested, we have now added Bakarat et al. 2013, and Seneviratne et al. 2016 in our updated search and analyses. However, we had to exclude Oliveiria Melo et al. 2012 because it did not fulfill our inclusion criteria. Specifically, it did not report the diagnosis of GDM as one of its outcomes.

Submitted filename: Responses to Editor and Reviewers2.docx

Click here for additional data file.

26 Jul 2022

Effectiveness of exercise intervention during pregnancy on high-risk women for gestational diabetes mellitus prevention: a meta-analysis of published RCTs

PONE-D-21-21724R2

Dear Dr. Tatsioni,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Carsten Bogh Juhl, PhD

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #2: (No Response)

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #2: No

**********

28 Jul 2022

PONE-D-21-21724R2

Εffectiveness of exercise intervention during pregnancy on high-risk women for gestational diabetes mellitus prevention: a meta-analysis of published RCTs

Dear Dr. Tatsioni:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Carsten Bogh Juhl

Academic Editor

PLOS ONE