PLoS ONE
Public Library of Science
image
Assessing the impact of group antenatal care on gestational length in Rwanda: A cluster-randomized trial
DOI 10.1371/journal.pone.0246442 , Volume: 16 , Issue: 2 , Pages:
Article Type: research-article, Article History
Abstract

Background

Research on group antenatal care in low- and middle-income contexts suggests high acceptability and preliminary implementation success.

Methods

We studied the effect of group antenatal care on gestational age at birth among women in Rwanda, hypothesizing that participation would increase mean gestational length. For this unblinded cluster randomized trial, 36 health centers were pair-matched and randomized; half continued individual antenatal care (control), half implemented group antenatal care (intervention). Women who initiated antenatal care between May 2017 and December 2018 were invited to participate, and included in analyses if they presented before 24 weeks gestation, attended at least two visits, and their birth outcome was obtained. We used a generalized estimating equations model for analysis.

Findings

In total, 4091 women in 18 control clusters and 4752 women in 18 intervention clusters were included in the analysis. On average, women attended three total antenatal care visits. Gestational length was equivalent in the intervention and control groups (39.3 weeks (SD 1.6) and 39.3 weeks (SD 1.5)). There were no significant differences between groups in secondary outcomes except that more women in control sites attended postnatal care visits (40.1% versus 29.7%, p = 0.003) and more women in intervention sites attended at least three total antenatal care visits (80.7% versus 71.7%, p = 0.003). No harms were observed.

Interpretation

Group antenatal care did not result in a difference in gestational length between groups. This may be due to the low intervention dose. We suggest studies of both the effectiveness and costs of higher doses of group antenatal care among women at higher risk of preterm birth. We observed threats to group care due to facility staff shortages; we recommend studies in which antenatal care providers are exclusively allocated to group antenatal care during visits.

Trial registration

ClinicalTrials.gov NCT03154177

Sayinzoga, Lundeen, Musange, Butrick, Nzeyimana, Murindahabi, Azman-Firdaus, Sloan, Benitez, Phillips, Ghosh, Walker, and Adu-Afarwuah: Assessing the impact of group antenatal care on gestational length in Rwanda: A cluster-randomized trial

Introduction

Background

In 2016, the World Health Organization recommendations on antenatal care for a positive pregnancy experience prioritized research on the individual outcomes and health systems effects of group antenatal care implementation [1]. Research to date on this alternative model of antenatal care provision shows mixed results. An individually randomized controlled trial (RCT) among women in the United States at high socio-demographic risk for preterm birth showed that participation in group antenatal care was associated with a significantly lower preterm birth rate (9.8%) than participation in only individual antenatal care (13.8%) [2]. A cluster RCT among a similar cohort of women found that group antenatal care reduced the rate of small-for-gestational-age infants, and increased gestational length among small-for-gestational-age infants [3]. A meta-analysis including these studies, two other RCTs and ten observational studies found no differences in preterm birth or low birth weight [4]. However, when the authors performed a sub-group analysis by race/ethnicity limited to the two highest-quality studies, the preterm birth rate was significantly lower among African-American women who participated in group antenatal care (8.0%) compared to African-American women who participated in individual antenatal care (11.1%). Group antenatal care is hypothesized to positively impact preterm birth rates and other outcomes among women at elevated psychosocial risk due to three main features of the model: 1) greater social support between women who are linked via the group; 2) more total antenatal care-associated time spent in educational activities in facilitated group discussions; and 3) attention to key elements of person-centered care, including respect and safety, empowerment, and participation [57]. These elements create a more positive pregnancy care experience which may encourage antenatal care attendance and thus create additional opportunity for risk assessment by providers.

Group antenatal care has been described in several low- and middle-income country contexts. A prospective cohort trial in Ghana reported significantly higher health literacy among women who participated in group antenatal care [8]. A pilot study in Tanzania and Malawi reported feasibility, acceptability and a significant increase in attendance at five antenatal care visits, as well as significantly more satisfaction with care [9, 10]. A cluster RCT of 20 facilities in Nigeria and 20 facilities in Kenya reported that group antenatal care was associated with a significantly higher rate of facility birth in Nigeria, but not in Kenya and an increase in attendance at four antenatal care visits in both countries [11]. No study completed in a low- or middle-income country context has yet reported on the effect of group antenatal care on gestational length, preterm birth, or low birth weight.

Objectives

The Preterm Birth Initiative-Rwanda is a partnership between investigators at the University of Rwanda and University of California, San Francisco and national health system implementors at the Rwanda Biomedical Center and Ministry of Health. Intrigued by lower rates of preterm birth among high-risk American women who participated in group antenatal care, the Preterm Birth Initiative-Rwanda aimed to test the primary hypothesis that Rwandan women receiving antenatal care at health centers that offer group antenatal care would experience increased gestational length compared to women receiving antenatal care at health centers that provide the standard, individual model of care. Stakeholders were also interested in the effect of group antenatal care on attendance, risk identification and other health outcomes.

Context

Rwanda’s national maternity care system provided an excellent opportunity to test this innovative service delivery model due to its community capacity, cultural foundations in community-based decision-making and cooperation, and extant, longitudinal antenatal care registers. During this study period, 2017–2018, Rwanda’s national guidelines prescribed that each childbearing woman be offered four focused antenatal visits (according to WHO recommendations prior to 2016) and four postnatal care visits (within 24 hours, 2–3 days,7–14 days, and 42 days of infant life). Currently, national guidelines are under revision to align with the 2016 recommendations [1].

In Rwanda, routine antenatal care is provided in government-system health centers staffed entirely by nurses and midwives. Women are required to pay a fee at the time of each antenatal visit and facility birth care. A community-based insurance scheme is available to all Rwandan families, which decreases but does not eliminate the cost of these services. Virtually all (99%) pregnant women attend at least one antenatal visit, while only 44% attend the four antenatal visits [12]. Antenatal care providers participate in a performance-based incentive program that rewards them by the proportion of pregnant women who enroll in antenatal care before 16 weeks and attend four antenatal visits following the focused antenatal schedule. While gestational age or gestational length assessments are challenging as early ultrasound is not routinely available in Rwandan health centers, 56% of women report they enter antenatal care before 16 weeks of pregnancy based on the last menstrual period [12]. 91% of births occur in a facility such as a health center or district hospital. About 19% of newborns and 43% of women receive a postnatal assessment within the first two days after birth.

This article reports the primary and secondary outcomes of this cluster RCT, including gestational length, mortality among preterm neonates, attendance at four antenatal care visits, attendance at the first antenatal care visit before 14 completed weeks gestation, attendance at a six-week postnatal care visit at a health center, identification of women as at high risk at any antenatal care visit, and caesarean section rates among enrolled women. Results of other outcomes, including women’s and providers’ experiences of group antenatal care, are reported elsewhere [1315].

Methods

Trial design

To test the primary hypothesis that group antenatal care would extend gestational length, the Preterm Birth Initiative-Rwanda designed a cluster RCT in which a cluster was defined as a health center and the population served in its catchment area. We chose a clustered design because Rwandan stakeholders preferred to offer all women at each health center the same model of care. Clusters were pair-matched and one of each pair was assigned to continue individual antenatal visits while the other was assigned to provide group antenatal visits.

At half of the health centers included in this study we implemented community-based urine pregnancy test by community health workers and basic obstetric ultrasound by nurses and midwives to strengthen gestational age assessment and assess whether these interventions affected the secondary outcomes of attendance at four antenatal care visits and initiation of antenatal care before 14 completed weeks. Health center pairs were matched to similar pairs to ensure balance, and one pair from each quadruple was assigned the additional intervention of basic obstetric ultrasound by nurses and midwives at the health center and community-based urine pregnancy tests administered by community health workers in the catchment area.

The intervention group consisted of all health centers providing group antenatal visits; half of these health centers also provided basic obstetric ultrasound and pregnancy testing at community level. The control group consisted of all health centers providing standard individual antenatal visits, half of which also provided basic obstetric ultrasound and pregnancy testing at community level.

Ethical approval

Ethical approval was granted by the Rwanda National Ethics Committee (No.0034/RNE/2017) and University of California, San Francisco Institutional Review Board (16–21177). Data were collected on women aged 15 and older presenting for antenatal care at the 36 study health care centers who provided written informed consent between May 25, 2017 and December 31, 2018. This study was permitted by the Rwanda National Ethics Committee to waive parental consent for pregnant minors ages 15 and older. No pregnant adolescents younger than 15 years were documented to have been invited to participate in the study.

Participant selection and recruitment

All pregnant women presenting for their first antenatal visit received standard individual care from a provider. Providers invited women to participate in the study and attend future antenatal visits at the study site according to the Rwanda focused antenatal care schedule. At sites randomized to group antenatal care, providers and study staff invited women to participate in group antenatal care; those who declined continued to receive individual care at the study site. After the antenatal care provider estimated the woman’s due date based on the last menstrual period and symphysis-fundal height, study staff assigned the woman to a group of eight to 12 women with similar due dates (preferably within a two-week gestational-age period, with an upper limit of 4 weeks difference). Once the woman was assigned to a group, she was invited to return for three subsequent scheduled group antenatal visits at eight-week intervals, starting at 22–24 weeks gestation, and one postnatal group visit. While group antenatal services were offered to all women at group care facilities, only women who consented to study participation, presented for the first antenatal care visit before 24 completed weeks gestation, attended at least two antenatal care visits during pregnancy, and whose birth outcomes were discoverable by study staff were eligible and included in the analyses.

The primary analysis included only those women with a gestational length between 24 weeks and 43 weeks, documented by the birth care provider in the birth facility’s maternity register. We included mother-infant units in the primary analysis if the infant’s birth weight fell between the Intergrowth-21st Project’s upper and lower centiles, by sex and gestational length [16]. The Intergrowth-21st Project provides international standards for female and male infants between the 3rd and 97th centiles by gestational length.

Study interventions

This study included one primary (group antenatal care) and two secondary (basic obstetric ultrasound and urine pregnancy tests in the community) interventions. These interventions are described below, and more details are available in a separate publication [17].

Group antenatal care and postnatal care

Each group antenatal visit occurred in a room accommodating eight to 12 pregnant women, one antenatal provider and one community health worker. During the first half of these two-hour sessions, antenatal care providers met with each woman in a semi-private area of the room for brief individual assessments while the group of women socialized and participated in health assessment activities such as weighing one another and taking each other's blood pressure readings with an electronic cuff, under the supervision of the community health worker. During the second hour of the session, the provider and community health worker co-facilitated a discussion of health-related topics that aligned with the stage of pregnancy and the health issues of highest importance at that time. The full list of visit timing and topics at each visit are described elsewhere [17]. Providers and community health workers elicited concerns and questions from women and encouraged group care participants to share knowledge and support with one another.

The Rwanda group antenatal care model was customized by a Technical Working Group convened for this study, composed of representatives from maternal-child health stakeholder organizations in Rwanda [18]. During the study period, the Rwanda Ministry of Health recommended that the group care curriculum follow the four focused antenatal visits model and no more than four antenatal visits per woman could be accommodated by the health system. The Technical Working Group hoped that the social support fostered among women in the same antenatal group would continue into the postnatal period and motivate women to seek care; for this reason, a postnatal group visit was included in the model even though a postnatal visit was not expected to impact the primary outcome (gestational length). Experienced group antenatal care providers from the United States prepared six Rwandan providers with five days of training. These Rwandan providers then trained three antenatal care providers at each health center randomized to group antenatal care, with a three-day group care training session. The Rwandan trainers continued to provide the 18 group antenatal care clusters with targeted training, supervision, and mentoring throughout the study.

Basic obstetric ultrasound by nurses and midwives at the health center

The Preterm Birth initiative-Rwanda provided one ultrasound machine and 10 days of training for three antenatal care providers per site to the 18 clusters randomized to implement basic obstetric ultrasound. These clusters were asked to conduct a screening ultrasound examination for each woman on the day of her first antenatal visit, or soon after. The Rwanda Society of Radiologist trained these new ultrasound providers who subsequently received mentoring and supervisory visits from radiographers from the nearest district hospital.

Urine pregnancy testing in the community

Community health workers associated with health centers randomized to urine pregnancy tests in the community underwent an eight-hour training and were provided with urine pregnancy test kits. They were instructed to refer women with a positive pregnancy test to the health center for antenatal care services. The Rwanda Biomedical Center trained these community health workers, and each health center’s community health worker supervisor supervised the community health workers in the catchment area.

Outcomes

The primary outcome was specified with a testable hypothesis a priori during clinical trial registration (clinicaltrials.gov NCT03154177). The primary study hypothesis was that among women who presented for antenatal care at <24 weeks gestation and attended ≥2 antenatal visits documented at the site of study enrollment, antenatal care exposure at sites that offered group antenatal visits would increase mean gestational length by .5 weeks (with a standard deviation no larger than 4.3 weeks) compared to antenatal care exposure at sites that offered standard antenatal visits. Gestational length was assigned by the birth care provider and recorded in the facility’s maternity register. Birth care providers used all available data to make this determination at the time of birth (i.e. last menstrual period, birth weight, infant maturity), but gestational length assignment was not standardized or monitored. Birth weight was recorded by birth providers before leaving the delivery room using an analog infant scale; this measurement was abstracted from the maternity register at each facility. Birth care providers were not informed about the study’s primary hypothesis nor primary outcome.

The secondary outcomes were adherence to the recommended four antenatal visits, gestational length at first antenatal visit, incidence of preterm birth, proportion of women delivering by caesarean section, proportion identified as at risk during antenatal care, and adherence to six-week postnatal visit. We intended to examine newborn morbidities but our data sources were not adequate to do so. Antenatal visits were counted in the longitudinal antenatal register at each facility. Gestational length at the time of the first antenatal visit was documented by the antenatal provider in the facility antenatal register. Preterm birth was assigned to any infant whose gestational length was recorded in the maternity register as less than 37 weeks. We also examined the effect of group antenatal care on mode of birth and the effect of basic obstetric ultrasound and community-level urine pregnancy tests on antenatal care attendance. Gestational length at the time of the first antenatal visit was documented by the antenatal care provider in the facility antenatal register. Antenatal visits were reported in the pre-existing longitudinal antenatal register at each facility. Postnatal visits were reported in the recently established postnatal register at each facility.

Data sources and collection

On the day women consented to study participation, immediately after the first antenatal visit, all study participants completed an enrollment questionnaire about socio-economic and health history, which was administered by a health provider or study staff member. Data were entered directly on mobile tablets by data collectors into an encrypted, web-based system called Research Electronic Data Capture (REDCap) platform [19].

Study staff abstracted all data from standard Ministry of Health primary (paper) data sources located in the study facilities. These included antenatal, birth, neonatal, and postnatal registers, and individual longitudinal antenatal and postnatal records. Birth registers were reviewed at all 36 health center study sites and the six district hospitals to which those health centers referred women for higher-level care. Study staff were instructed to extract newborn morbidities from health center data drawn from community health worker SMS reports, but in practice these data were not available.

The study initially created and used a multisite Research Electronic Data Capture database. However, due to poor connectivity and the quantity of data collected per participant (367 variables), the process of syncing or uploading data from the tablets to the server was frequently disrupted. The disruption led to many duplications and erroneous linkage of study events. Upon discovery of this error, the study team substantially revised the database and conducted a thorough process to rectify any errors and retain the most complete record for each patient. Unrectified records were excluded from the final datasets. The electronic data were maintained on secure systems with access limited to the principal investigators, study epidemiologist, and designated study staff. The data were converted into SPSS for Windows version 25 [20] and STATA SE version 16 [21] for analyses.

Sample size

At the design stage of this trial, we assumed an ICC of 0.01, and a possible effect size of half a week in gestational length based on a Cochrane review of conventional versus group antenatal care [22]. We assumed a loss-to-follow-up rate of 30%, based on the assumptions that 10–15% of all pregnancies end in miscarriage and 15–20% of women might deliver at another location. We calculated that for a two-tailed test, α = 0.05, 1-β = 80%, and a balanced 1:1 ratio of intervention (group antenatal care) to control (standard antenatal care) study participants, a minimum sample of n = 1,163 eligible women per study group (intervention and control) was required. The sample size was increased to account for a cluster design effect of 3.21 (cluster randomization to study group by health center rather than by individual) and to account for a loss to follow-up rate of up to 50%, for a total required sample size per study group of 3,640, or an average of 202 women per health center at 36 health centers [23]. As descriptive outcomes, the secondary outcomes were assessed in the available sample without hypothesis testing or multiple comparisons adjustment.

Site selection

Five of 30 districts in Rwanda, including Burera, Bugesera, Nyamasheke, Rubavu, and Nyarugenge, were selected by the Ministry of Health as locations for this cluster RCT based on their service capacity and need. To gather information that would inform study site selection, data collectors visited all 55 public health centers in the study districts and interviewed staff about the facility itself, human and material resources, and client volumes. We selected 36 health centers for inclusion that 1) had historical volumes of women that made organizing groups feasible (35–125 births at the health center per month); 2) reported at least two antenatal care providers per day when antenatal care was offered, to increase the likelihood that one antenatal care provider could be exclusively allocated to scheduled group visits; and 3) had a room sufficiently large for group visits.

Randomization

To optimize study group comparability, 36 selected health centers were pair-matched. First, the study statistician applied a non-bipartite matching algorithm in R, using the nbpMatching package (available from: https://cran.r-project.org/web/packages/nbpMatching/index.html), including health center-specific data on the monthly number of women registering for antenatal care, monthly number of births, proportion of first antenatal visits completed before 16 weeks gestation, and availability of screening tests as a composite. Thirty sites were matched with the non-bipartite matching algorithm [24] with the remaining six presented as multiple options, which were then reviewed by the study team that finalized matching based on the monthly number of women registering for antenatal care, rural/urban designation and distance to the nearest hospital. After matching was complete, pairs were further matched to quadruples. Then one site in each pair was assigned to group antenatal care and the other to standard antenatal care using random selection in R software [25]. One pair of each quadruple was similarly randomly selected for implementation of basic obstetric ultrasound and urine pregnancy testing in the community.

No allocation concealment was used. The Ministry of Health notified the heads of the health centers of their selection status; all selected health centers agreed to participate.

Statistical analyses

Cleaning (range and logic) checks were applied, and eligibility and critical outcomes data requiring further cleaning were sent to designated field staff to review and resolve discrepancies. We conducted individual level bivariate analyses stratified by study group to assess study group comparability.

We compared the control group (individual antenatal visits) with the intervention group (group antenatal visits) in the primary analysis. The primary analysis was conducted using gestational length as recorded by the maternity providers after conducting sensitivity analyses to determine if any other gestational length variable or computation would produce results that were more plausible for correct gestational length classification. We relied on the Intergrowth 21st Project’s international standards for birth weight by sex and gestational length to determine the plausibility of correct GA classification, excluding infants with birth weights outside the upper and lower centiles.

Treatment effect was estimated using linear and logistic regression generalized estimating equations with robust variance estimation to account for clustering of births within facility and to adjust for pairing of facilities. We used exchangeable correlation structure and conducted an individual level analysis. We developed a Directed Acyclic Graph to identify potential confounders and mediators for which data were available. In descriptive analyses, we investigated the factors on which the intervention and control arms were statistically significantly different. The final model was adjusted for pairing and clustering only. In sensitivity analyses, we additionally adjusted for potential confounders or mediators to demonstrate that the results were unaffected by residual confounding that may have resulted due to unbalanced study arms. Analogous analyses were conducted to assess continuous and categorical secondary outcomes. Significance tests were two-tailed at the 5% level. Analyses were conducted using SPSS v25 [17, 20] and STATA v16 [21].

Results

The Consolidated Standards of Reporting Trials (CONSORT) diagram is presented as Fig 1 showing that 25,258 women consented to participate in the study. Of those, 9,420 were excluded because they presented after 24 weeks or did not attend two antenatal visits at the facility or for other reasons. The paired randomization by health center produced a nearly equal number of women presenting for antenatal care across the study groups. The average number of women recruited per health center was similar between study groups (mean 695, range 146–1015 in control clusters and mean 708, range 194–1090 in intervention clusters). In total, 3,918 women were lost to follow-up (i.e., no birth outcome captured), with similar numbers lost in each study group. Prior to analysis, 3,077 women were eliminated because of missing or implausible outcome data. Thus, 8,843 women were included in the primary analysis because they had gestational length outcomes between 24–43 weeks documented by March 31, 2019 and the infants’ birth weights were plausible using the upper and lower centiles of the Intergrowth 21st Project’s standards. There were 4,091 eligible mother-baby units in the control group and 4,752 eligible mother-baby units in the intervention group.

Consolidated Standards of Reporting Trials (CONSORT) diagram for the Preterm Birth Initiative-Rwanda trial of group antenatal care.
Fig 1
Consolidated Standards of Reporting Trials (CONSORT) diagram for the Preterm Birth Initiative-Rwanda trial of group antenatal care.

Facility characteristics

Based on the design and data obtained for matching facilities before the intervention began, facilities randomized to the control and intervention conditions were similar (Table 1).

Table 1
Baseline facility characteristics by study group.
Control (n = 18)Intervention (n = 18)
Mean (range) / %Mean (range) / %
Births per month at the facility57 (32–123)50 (29–99)
Women per month who enroll for antenatal care at the facility98 (54–183)97 (58–248)
Mean proportion of pregnant women who attended the first antenatal care at gestational age less than 16 completed weeks (range)45% (21%-96%)53% (25%-88%)
Basic antenatal screening tests available, out of 54.7 (3–5)4.5 (3–5)
Proportion of facilities considered to be in a rural setting83%72%
Distance to referral hospital, in kilometers31 (10–60)32 (5–80)

Participant characteristics

While various differences between the study groups' participant characteristics were statistically significant, we do not consider these differences to be practically or clinically significant (Table 2). Fewer women in the intervention condition had health insurance, but among women in the lowest income category, more were enrolled in the community-based insurance system. Women in the intervention condition had higher levels of education, more were nulliparous, and more were short of stature (height <150 cm) or had a small middle-upper arm circumference (<21cm) when compared to women in the control condition. Fewer women in the intervention condition were over 35 years of age or smoked tobacco, while more women in the intervention condition were HIV-positive. Among multiparous women, more in the intervention condition reported a history of preterm birth or stillbirth. These differences were incorporated into the adjusted analyses of the primary and secondary outcomes.

Table 2
Women’s characteristics at the first antenatal visit, non-missing observations.
Characteristics of WomenControlIntervention
AgeN%N%
<18771.0750.9
18–35616581.9696185.6
>35128317107013.2
Household socio-economic status: “Ubudehe” category *
Category 1 (poorest)89218·394319.1
Category 2198640·7190438.6
Category 3175435.9162032.9
Category 4 (richest)30.160.1
I don’t know481.01603.2
None1984.02976·0
Currently has health insurance
No4889.497919.8
Yes469290.6397280.2
Education level
None6919.24986.1
Some primary310041.2290035.3
Completed primary195726.0239829.2
Some secondary99513.2129015.7
Completed secondary6258.383710.2
Some college or university650.91281.6
Completed college or university951.31581.9
Work outside her home
Unemployed6508.693811.4
Professional/technical/managerial1151.51471.8
Clerical50.140.0
Sales and services4335.74395.3
Skilled manual140.2320.4
Unskilled manual1872.52483.0
Domestic service1431.96097.4
Agriculture592378.2568368.8
Other610.81001.2
Missing480.6590.7
Cooking fuel at home
Wood & Charcoal744398.6803997.9
Electricity, Kerosene, Gas1041.41762.1
Risk factors for preterm birth
Height < 150 cm2393.23754.6
Middle-upper arm circumference < 21 cm1642.23544.3
Previously known human immunodeficiency virus status positive1021.71361.9
Presence of a household smoker3724.93264.0
Nulliparous187325.5247230.2
History of diabetes210.3280.3
History of hypertension370.5350.4
History of stillbirth881.61372.4
History of infant with low birth weight1842.41321.6
History of preterm birth160.3380.7
Elevated blood pressure today60.1150.2
Anemia140.2230.3
Multiple gestation diagnosed today1922.51521.8
None reported698592.2765192.6
*An Ubedehe category is assigned to each household. Local community members at the cell level are required to gather community members together and, with the help of Ubudehe facilitators/trainers, identify and place community members into different economic categories, ranging from the poorest households (lowest category) to the richest households (highest category) [26].

Outcomes

Primary outcome

The primary outcome, gestational length, was identical in the intervention and control groups (39.3 weeks (SE 0.2) and 39.3 weeks (SE 0.2), respectively; Table 3). The intra-cluster correlation coefficient was 0.00063. When we examined a subset of women in both study conditions (group antenatal care versus standard antenatal care) who also received an ultrasound examination between 6 and 22 weeks gestation and we calculated gestational length using that ultrasound data, the mean gestational length remained similar between groups (39.8 weeks (SE 0.03) versus 39.5 weeks (SE 0.03), respectively; S1 Table).

Table 3
Distribution of the primary and secondary outcomes by study groups and the effect of the intervention on these outcomes.
ControlIntervention
Primary Outcomen1/n2Mean (SE)n1/n2Mean (SE)β (95% Confidence Interval)aP-value
Gestational length (in weeks)409139.3 (0.02)475239.3 (0.02)-0.07 (-0·18, 0·04)0.24
Secondary outcomesn1/n2%n1/n2%Odds ratio (95% Confidence Interval) aP-value
Preterm birth146/40913.6177/47523.71.06 (0·73, 1.53)0.77
Cesarean birth444/408410.9573/474512.11.00 (0·78, 1.28)0.98
Proportion of women who attended at least 4 antenatal visits2654/757935.03478/825942.11.20 (0·86, 1.68)0.29
Proportion of these women who attended 1st antenatal visit before 14 completed weeks gestation3148/751641.93040/822337.00.86(0.61, 1.21)0.39
Mortality among preterm neonates (among all preterm babies)14/1469.613/1777.3b
Attendance at postnatal visits at 6 weeks (among all eligible women with delivery outcome)1675/409140.11413/475229.70.52 (0.33, 0.80)0.003
Women identified at high risk (among all eligible women, including any high risk in pregnancy)c623/409115.2729/475215.30.98 (0.71, 1.35)0.89
a Models are adjusted for pairing and clustering of births within facilities.
b Sufficient observations are not available for the models to fit.
c Any high risk in pregnancy includes history of low birth weight, preterm birth, stillbirth, neonatal death, or detection of anemia, edema, proteinuria, not gaining weight, high bp, multiples, abnormal lie after 32 weeks or other provider initiated referral to higher level care.
n1 = Numerator for the specific category in the control/intervention arm
n2 = Total number of non-missing observations for the respective variable in the control/intervention arm.

Secondary outcomes

Secondary outcomes among women in control and intervention groups are included in Table 3. There were no significant differences between study groups in preterm birth or mode of birth by Cesarean. There was no difference between groups in attendance at four antenatal visits and first antenatal visit before 14 completed weeks of pregnancy. Mortality among preterm neonates was slightly lower in the intervention group, however the number of data points in this category was insufficient to perform inferential statistics. Attendance at the six-week postnatal visit was higher in the control group than the intervention group. There was no difference in the proportion of women identified as having a risk factor during pregnancy between the two study groups.

Exploratory analysis revealed no difference in low birth weight between groups (S3 Table). Further adjusted analyses for factors associated with preterm birth and low birth weight did not change the results, except that multiple gestation and maternal weight less than 45 kilograms or over 80 kilograms were associated with higher low birth weight in the intervention group.

Among women with subjective or measurable risk factors for poor perinatal outcomes (“high-risk women”), the odds ratio for low birth weight was significantly higher among women who received antenatal care at clusters randomized to group antenatal care compared to those at clusters randomized to standard antenatal care, but only among women with maternal pre-pregnant weight less than 45 kilograms or more than 80 kilograms or with multiple pregnancy (S2 Table). There were no other differences in outcomes for “high-risk women” between study conditions with respect to gestational length, preterm birth, or mode of delivery.

Exploratory analysis to better understand attendance patterns showed that more women who attended antenatal care at facilities randomized to group care attended at least three total antenatal visits compared to women in the control group (80.7% versus 71.7%, p = 0.003) and that the mean number of antenatal visits attended was slightly higher (3.19 versus 3.05, p = 0.05; S3 Table).

There was no difference in gestational age at entry to antenatal care or attendance at four antenatal visits among women who attended antenatal care at sites that implemented urine pregnancy tests in the community and basic obstetric ultrasound at the health center compared to those without these interventions (Table 4).

Table 4
Antenatal care attendance among eligible women who presented at <24 weeks and attended at least two antenatal visits, comparing women from clusters randomized to Urine Pregnancy Testing (UPT) and ultrasound (US) to women from clusters randomized to no UPT by community health workers nor US at the health center level.
Yes UPT and USNo UPT nor USOdds ratio (95% Confidence Interval) *p-value*
n1/n2%n1/n2%
Proportion of women who attended at least 4 antenatal visits2876/710440.53256/873437.31.07(0.85 1.35)0.57
Proportion of these women who attended 1st antenatal visit before 14 completed weeks gestation2899/703741.23289/870237.81.07 (0.86, 1.34)0.54

Discussion

In this cohort of 8,843 women who attended an average of three antenatal visits, we did not find a difference in mean gestational length between study groups. Fewer than 4% of births were classified as preterm, but this result is substantially lower than other estimates for Rwanda [27] and is likely influenced by multiple biases inherent in gestational length misclassification. We used gestational length at birth as documented by the maternity provider in the facility register as our endpoint. This variable, when compared to all the gestational length-related data items available in our data set, resulted in the highest proportion of infants whose birthweights were plausible by Intergrowth 21st Project’s standards by infant sex. However, by applying these birth weight standards we excluded 26% of infants that had been classified as preterm because their birth weights were deemed implausible. This exclusion also resulted in fewer than 3% of infants with low birth weight, which is also inconsistent with reports from Rwanda [12]. These misclassification issues may have increased the similarity of gestational length at birth across groups.

In addition, the gestational length variable was 28% more likely to be missing for infants not born in the health center at which the mother attended antenatal visits; infants born outside the health center were born at the referral hospital, private facilities, or at home and may have disproportionately included preterm infants. Women with pregnancy risk factors identified by providers at the first antenatal visit were routinely referred to the district hospital for surveillance, and these high-risk women were less likely to continue antenatal care at the health center; this may have skewed our sample toward a more low-risk population with lower incidence of preterm birth and low birth weight.

To our knowledge, this is the largest-scale implementation of group antenatal care in an low- middle-income country context to date and we relied on existing register data documenting maternal and infant outcomes. This intervention did not include any additional provider staff, patient incentives or extra communications to women enrolled in the study.

The intervention dose was limited to about two group antenatal visits (three total antenatal visits), on average, among women in the intervention clusters. We hoped eligible women would attend three group antenatal visits during pregnancy according to the focused four-visit antenatal care model advocated by WHO before 2016, but it appears that the barriers to antenatal care attendance were greater than the potential appeal of the group care model. We hypothesize that two main barriers influenced women’s antenatal care attendance. First, women who presented for antenatal care at health centers were expected to enroll their families in the community-based insurance program and pay an annual premium (according to their income level) and co-pay at each visit according to their income level. If women do not enroll in community-based health insurance, they are obliged to cover the total cost of all facility services. Women and their families are required to present vital records and proof of income (Ubudehe category) to enroll in the insurance program. Before each antenatal visit, women are required to make a co-payment, which many of them find difficult to pay and thus could be a barrier to antenatal care attendance [14]. The second barrier to antenatal care attendance may paradoxically be related to the performance-based incentive program in place at these health centers. Antenatal care providers are financially incentivized when women in the catchment area present for the first antenatal visit before 16 weeks and then attend the following three antenatal visits within the gestational age ranges recommended by the focused antenatal care model. Visits outside the narrow gestational age ranges suggested in the focused four-visit antenatal care model were either not allowed or when they actually occurred they were not documented.

As reported in Table 3, we did not find a difference in gestational length among Rwandan women who attended an average of three total antenatal visits (one standard visit and two group visits in the intervention condition) compared to an average of three total standard antenatal visits (control condition). The intervention dose may have been too low to have an effect on health outcomes. A 2016 cluster RCT of group antenatal care in the United States reported that attendance at five or more group visits (of ten group visits offered in the intervention condition) was associated with improvements in all measured outcomes [3]. The minimum therapeutic dose is likely different in each context, but it appears that a dose of at least five group antenatal visits should be studied when health outcome endpoints are of interest. Grenier et al reported that Kenyan and Nigerian women who received care at facilities randomized to group antenatal care attended a total of six antenatal visits (median) [11]. We propose that future research of group antenatal care in low- and middle-income country contexts should study the effects of at least five group antenatal visits, within the WHO 2016 antenatal care framework of at least eight total antenatal contacts offered to pregnant women.

While there was a statistically significant difference regarding attendance at the six-week postnatal visit, it favored the control rather than the intervention group. There may be several potential reasons for this. First, a six-week postnatal care visit at facilities was not as clearly established in Rwanda as antenatal care patterns were. Thus, we are unclear whether postnatal measurements in all facilities were practically measuring the same thing, or whether women in control visits may have had immunization visits for their infants counted as postnatal care visits whilst group care sites only counted group visits. Second, women themselves may have valued a brief immunization focused visit rather a longer group session. Therefore, as group postnatal visits were set for a specific date from the time of the first antenatal visit, it is possible that a broader distribution of actual delivery dates than initially expected made postnatal care visit timing inconvenient or inappropriate for some women. We recommend further research in this area for others implementing group postnatal visits.

The change to group antenatal care provision at the 18 intervention health centers was disruptive, as any change will be. The most common concerns expressed by group antenatal care providers were: 1) the staffing model at these health centers often required that antenatal care providers also cover the maternity service, where women arrived in labor and needed immediate attention even if a group antenatal visit was in session; 2) both women and providers found it challenging to start the group antenatal visits at the scheduled time; and 3) the group care model required additional staff time and attention to create a group visit plan that would work, month to month, within the operations of the health center [13]. Without additional human resources to manage these challenges, antenatal care providers had to balance their enthusiasm for the group antenatal care model with the burden of additional tasks. We report in a separate publication that the group antenatal care model was delivered at approximately 80% fidelity across more than 2,600 documented visits [15]. The lowest-rated fidelity outcome was keeping to the intended time frame for each group visit, which was related to antenatal care staff being called to other services, especially labor and birth, when facilities were short-staffed.

Limitations

The variable we used for the primary outcome is documented by maternity providers in whole weeks, which limited the granularity of the analysis for differences between study groups. The generalizability of these results may be limited by the fact that our sample was restricted to women who registered for antenatal care before 24 completed weeks and attended at least two antenatal visits, which may represent a low-risk population of women in this context. We do not have outcome data for 24.7% of our cohort, which may also limit the generalizability of these findings; data transmission errors related to network connection problems contributed to this loss. We encountered substantial difficulties using a web-based data collection system in remote areas that had unstable connectivity. Finally, while use of facility data overall could be seen as a strength, we found that data collector access to newborn morbidity was not a reliable way to collect this information and were unable to complete analysis of this secondary outcome.

Conclusion

The group antenatal care model implemented in our study did not result in a difference in gestational length or preterm birth rate in the intervention group compared to the control group. In order to understand whether this intervention will improve health outcomes for other populations of women, we suggest follow-up studies of both the effectiveness and costs of higher doses of group antenatal care among women at a higher baseline risk of preterm birth and only in facilities in which antenatal care providers are exclusively and reliably allocated to group antenatal care provision during scheduled group visits.

Acknowledgements

First and foremost, we would like to acknowledge the women and their infants who provided consent to participate in this study. We would also like to thank the providers and community health workers who participated in this study, whose daily work is humbling to us all. We are grateful to the Rwanda Ministry of Health and Rwanda Biomedical Center and the head of health center at each of the 36 study facilities. We are grateful for the many people who contributed to this project over time, including Janine Condo, Jean-Baptiste Byiringiro, Catherine Mugeni, Evodia Dushimimana, Andrew Muhire, Vedaste Ndahindwa, Lauriane Nyiraneza, Nicole Santos, Caroline Kusi, Yvonne Delphine Nsaba Uwera, Antoinette Kambogo, Angele Musabyimana, Alice Umukunzi, Olive Tengera, Athanasie Mbuguje, Grace Liu, Matthew Meeks, Hannah Park, Mona Sterling, Fidens Dusabeyezu, and Wenjing Zheng. We thank our program officers France Donnay, Janna Patterson, Jerker Liljestrand, and Manpreet Singh for their guidance, encouragement, and support. We are grateful for the members of the Preterm Birth Initiative East Africa External Advisory Committee, our colleagues at the California Preterm Birth Initiative, and the members of our joint Strategic Advisory Board.

References

WHO recommendations on antenatal care for a positive pregnancy experience.

IckovicsJR, KershawTS, WestdahlC, MagriplesU, MasseyZ, ReynoldsH, et al Group prenatal care and perinatal outcomes: a randomized controlled trial. Obs Gynecol. 2007;110(2 Pt 1):3309. doi: 10.1097/01.AOG.0000275284.24298.23

IckovicsJR, EarnshawV, LewisJB, KershawTS, MagriplesU, StaskoE, et al Cluster randomized controlled trial of group prenatal care: Perinatal outcomes among adolescents in New York city health centers. Am J Public Health. 2016;106(2):35965. doi: 10.2105/AJPH.2015.302960

MazzoniSE, CarterEB. Group prenatal care. American Journal of Obstetrics and Gynecology. 2017 doi: 10.1016/j.ajog.2017.02.006

RisingS. Centering Pregnancy An Interdisciplinary Model of Empowerment. J Nurse Midwifery. 1998;43(1):4654. doi: 10.1016/s0091-2182(97)00117-1

ManantA, DodgsonJE. CenteringPregnancy: An Integrative Literature Review [Internet]. Vol. 56, Journal of Midwifery and Women’s Health. J Midwifery Womens Health; 2011 [cited 2020 Aug 19]. p. 94102. Available from: https://pubmed.ncbi.nlm.nih.gov/21429072/ doi: 10.1111/j.1542-2011.2010.00021.x

ThielenK. Exploring the Group Prenatal Care Model: A Critical Review of the Literature. J Perinat Educ [Internet]. 2012 [cited 2020 Aug 19];21(4):20918. Available from: /pmc/articles/PMC3489125/?report = abstract doi: 10.1891/1058-1243.21.4.209

LoriJR, Ofosu-DarkwahH, BoydCJ, BanerjeeT, AdanuRMK. Improving health literacy through group antenatal care: a prospective cohort study. BMC Pregnancy Childbirth. 2017;17(1):228 doi: 10.1186/s12884-017-1414-5

PatilCL, KlimaCS, SteffenAD, LeshabariSC, PaulsH, NorrK. Implementation challenges and outcomes of a randomized controlled pilot study of a group prenatal care model in Malawi and Tanzania. Int J Gynecol Obstet. 2017;139(3):2906. doi: 10.1002/ijgo.12324

10 

PatilCL, KlimaCS, LeshabariSC, SteffenAD, PaulsH, McGownM, et al Randomized controlled pilot of a group antenatal care model and the sociodemographic factors associated with pregnancy-related empowerment in sub-Saharan Africa. BMC Pregnancy Childbirth. 2017; doi: 10.1186/s12884-017-1493-3

11 

GrenierL, SuhowatskyS, KabueMM, NoguchiLM, MohanD, KarnadSR, et al Impact of group antenatal care (G-ANC) versus individual antenatal care (ANC) on quality of care, ANC attendance and facility-based delivery: A pragmatic cluster-randomized controlled trial in Kenya and Nigeria. PLoS One. 2019; doi: 10.1371/journal.pone.0222177

12 

National Institute of Statistics of Rwanda, Ministry of Finance and Economic Planning, Ministry of Health TDPII. Rwanda Demographic and Health Survey, 2015–2016. Kigali, Rwanda: NISR, MOH, and ICF International; 2016.

13 

LundeenT, MusangeS, AzmanH, NzeyimanaD, MurindahabiN, ButrickE, et al Nurses’ and midwives’ experiences of providing group antenatal and postnatal care at 18 health centers in Rwanda: A mixed methods study. PLoS One. 2019; doi: 10.1371/journal.pone.0219471

14 

MusabyimanaA, LundeenT, ButrickE, SayinzogaF, RwabufigiriBN, WalkerD, et al Before and after implementation of group antenatal care in Rwanda: A qualitative study of women’s experiences. Reprod Health. 2019;

15 

ButrickE, LundeenT, PhillipsBS, TengeraO, KambogoA, UweraYDN, et al Model fidelity of group antenatal and postnatal care: a process analysis of the first implementation of this innovative service model by the Preterm Birth Initiative-Rwanda. Gates Open Res. 2020; doi: 10.12688/gatesopenres.13090.1

16 

VillarJ, IsmailLC, VictoraCG, OhumaEO, BertinoE, AltmanDG, et al International standards for newborn weight, length, and head circumference by gestational age and sex: The Newborn Cross-Sectional Study of the INTERGROWTH-21st Project. Lancet. 2014;

17 

MusangeSF, ButrickE, LundeenT, SantosN, Azman FirdausH, BenitezA, et al Group antenatal care versus standard antenatal care and effect on mean gestational age at birth in Rwanda: protocol for a cluster randomized controlled trial. Gates Open Res. 2019;3:1548 doi: 10.12688/gatesopenres.13053.1

18 

SayinzogaF, LundeenT, GakwerereM, ManziE, NsabaY, UmuzigaM, et al Use of a Facilitated Group Process to Design and Implement a Group Antenatal and Postnatal Care Program in Rwanda. J Midwifery Womens Heal. 2018;J Midwifer(Sept 25). doi: 10.1111/jmwh.12871

19 

HarrisPA, TaylorR, ThielkeR, PayneJ, GonzalezN, CondeJG. Research electronic data capture (REDCap)–A metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inf. 2009;42(2):37781. doi: 10.1016/j.jbi.2008.08.010

20 

IBM Corp. Released. IBM SPSS Statistics version 25.0. 2017.

21 

Press S. Stata Statistical Software: Release 16. StataCorp LLC. 2019;

22 

CatlingCJ, MedleyN, FoureurM, RyanC, LeapN, TeateA, et al Group versus conventional antenatal care for women. Cochrane Database Syst Rev [Internet]. 2015 2 4 [cited 2020 Aug 27];(2). Available from: http://doi.wiley.com/10.1002/14651858.CD007622.pub3

23 

HayesRJ, MoultonLH. Cluster randomised trials, second edition [Internet]. Cluster Randomised Trials, Second Edition CRC Press; 2017 [cited 2020 Aug 19]. 1–398 p. Available from: https://jhu.pure.elsevier.com/en/publications/cluster-randomised-trials-second-edition

24 

LuB, GreevyR, XuX, BeckC. Optimal nonbipartite matching and its statistical applications. Am Stat [Internet]. 2011 2 [cited 2020 Aug 20];65(1):2130. Available from: /pmc/articles/PMC3501247/?report = abstract doi: 10.1198/tast.2011.08294

25 

Team RC. R: A Language and Environment for Statistical Computing. Vienna, Austria. 2019;

26 

SimPN· K-WL· S. A study on the classification of households in Rwanda based on factor scores. Korean Data Inf Sci Soc [Internet]. 2018 3 31 [cited 2020 Aug 19];(2):54755. Available from: http://www.kdiss.org/journal/view.html?spage=547&volume=29&number=2

27 

ChawanpaiboonS, VogelJP, MollerA-B, LumbiganonP, PetzoldM, HoganD, et al Articles Global, regional, and national estimates of levels of preterm birth in 2014: a systematic review and modelling analysis. Lancet Glob Heal [Internet]. 2019 [cited 2020 Aug 19];7:e3746. Available from: www.thelancet.com/lancetgh

14 Jul 2020

PONE-D-20-10387

Assessing the impact of group antenatal care on gestational length in Rwanda: a cluster-randomized trial

PLOS ONE

Dear Dr. Lundeen,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Aug 28 2020 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

  • 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: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Seth Adu-Afarwuah

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

2. You indicated that you had ethical approval for your study.

In your Methods section, please ensure you have also stated whether you obtained consent from parents or guardians of the minors included in the study or whether the research ethics committee or IRB specifically waived the need for their consent, or whether minors with children are legally allowed to consent for themselves.

3. In your Data Availability statement, you have not specified where the minimal data set underlying the results described in your manuscript can be found. PLOS defines a study's minimal data set as the underlying data used to reach the conclusions drawn in the manuscript and any additional data required to replicate the reported study findings in their entirety. All PLOS journals require that the minimal data set be made fully available. For more information about our data policy, please see http://journals.plos.org/plosone/s/data-availability.

Upon re-submitting your revised manuscript, please upload your study’s minimal underlying data set as either Supporting Information files or to a stable, public repository and include the relevant URLs, DOIs, or accession numbers within your revised cover letter. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories. Any potentially identifying patient information must be fully anonymized.

Important: If there are ethical or legal restrictions to sharing your data publicly, please explain these restrictions in detail. Please see our guidelines for more information on what we consider unacceptable restrictions to publicly sharing data: http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions. Note that it is not acceptable for the authors to be the sole named individuals responsible for ensuring data access.

We will update your Data Availability statement to reflect the information you provide in your cover letter.

4. Please note that in order to use the direct billing option the corresponding author must be affiliated with the chosen institute. Please either amend your manuscript or remove this option (via Edit Submission).

5. We noted in your submission details that a portion of your manuscript may have been presented or published elsewhere:

'Table 1 was previously published in the Journal of Midwifery and Women's Health. Figure 1 was previously published in Gates Open Research. These publications are included as Related Manuscripts in this submission.'

Please clarify whether this publication was peer-reviewed and formally published.

If this work was previously peer-reviewed and published, in the cover letter please provide the reason that this work does not constitute dual publication and should be included in the current manuscript.

[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: Partly

Reviewer #2: Partly

Reviewer #3: Yes

**********

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

Reviewer #1: I Don't Know

Reviewer #2: No

Reviewer #3: 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

Reviewer #3: No

**********

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: Yes

Reviewer #3: 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: Thank you for the opportunity to read a well written paper on an important topic: addressing ANC and maternal/infant health in LMIC's/economically disadvantaged populations is very important. I do have some experience in healthcare in LMIC's in the African continent (but not Rwanda), cluster RCT's and GEE. I appreciate the complexity of the design, and the detailed explanations. I will focus more on technical issues related to design and analysis as this topic is not my clinical area of expertise.

1. This is a complicated design. I can not recall seeing an example of a cluster RCT with a factorial component. I remain unclear on aspects of your analysis. Usually a factorial design will report covariates for each factor, and an interaction estimate (between factors). The interaction effect is hopefully null, and then each factor effect can be reported independently. Was this done? Within a GEE, it is beyond my technical grasp how this would be setup given additional covariate for estimating random effects (intraclass correlations) as described in the manuscript. I think additional explanation is required, and perhaps supplemental analytical detail.

2. The pair matching is an additional complicated aspect of the design. As I read your methods, it was unclear to me if the 4 factors for matching were considered separately, or were considered as a composite. I think the 'range' of values which were considered a 'match' should be reported (e.g. 50%+/-5% was considered a match for proportion of first ANC visits, monthly delivery volume was 60 +/-10, et cetera). If matching variables were treated independently (and not as a composite) I'm not certain of the validity of reporting variables used in matching in Table 2 (unless to confirm no failure in matching) as variables used for matching must be within the specified parameters of the original matching unless there was a technical failure. I believe this would apply to the first 4 variables in Table 2.

3. There is a risk of selection bias given the loss to follow-up. It is reassuring that the relative proportions reported in Figure 2 are somewhat similar. Was there any analysis performed on characteristics between mothers who completed the study compared to those where mother-baby pair data was available?

4. Why was GL/GA not calculated based on the estimate from the first ANC rather than based at the time of delivery? Given entry into the study based on this estimate, and issues of blinding (although assessors were unblinded, it is an objective measure of GL/GA estimate if using the estimate from first ANC and date of delivery).

Minor comment

1. In the results section, I think more technically accurate language would be that participant characteristics were statistically significant, but were not considered to be practically or clinically significant (I'm not sure of a technical meaning for insubstantial). I'm not a fan of statistical testing of demographic features between treatment arms, and your study might be a good example of permitting the reader to draw their own conclusion as to if any of these features are practically significant (rather than being drawn to a p value that perhaps better reflects size of the study).

Reviewer #2: Abstract:

- since this was a cluster randomised trial, the reported findings should also have the number of clusters included in the analysis. You should report this alongside the numbers of women in each arm, e.g. "4091 women in X clusters and 4752 women in Y clusters at control and intervention facilities, respectively, were included..."

- you need to specify the statistics represented by the "±1·6" and "±1·5" - is this the range of values? If so say so; if it is another statistic other than the range, please remove the ± designation.

Objectives:

- the objectives make no mention of the ultrasound component of the intervention; indeed having read it, the description of the trial seems incomplete to me - it is a factorial cluster randomised trial testing the effect of group ANC with or without obstetric ultrasound at the health centre on gestational age at birth.

Methods:

- in study endpoints where it is indicated that the study hypothesised that group ANC would increase mean GA at birth by 0·5±4·3 weeks, you should specify the statistic represented by "±4·3"; if it is not a range, then the ± designation should not be used.

- for the sample size calculation, you should indicate the cluster design-related parameters by which the sample size was inflated by to account for this design, with a justification for the choice.

- please organise the sub-sections in the methods to follow the sequence recommended in the CONSORT guidelines

- individual level bivariate analyses stratified by study group using Chi-square and Student’s t-test statistics are not appropriate for cluster-randomised trials because they do not adjust for clustering. In any case, statistical tests comparing groups in terms of baseline characteristics are not appropriate for randomised controlled trials.

- adjusting variables in a cRCT should be specified a priori and not based on post-hoc statistical considerations as described in the last paragraph of the methods.

Results:

- the CONSORT diagram is not appropriate for a cluster RCT; it should include the numbers of clusters at each stage and the mean or median number of women per cluster with the SD or IQR.

- the statistical tests comparing groups in table 3 are not appropriate in a cRCT

- avoid the ± designation all through the results and specify what the reported statistic is, whether range, SD or SE. For continuous outcomes, the mean and SD should be reported in the descriptive tables, and the mean and SE for inference (e.g. in Table 4, Table 5, Table 6, and the results of primary and secondary outcomes).

- see comment above about adjusting variables in randomised controlled trials.

Reviewer #3: This is an important paper summarising the results of a large cluster randomised controlled trial (RCT) assessing the effects of group antenatal care among women in Rwanda on child gestational length. The study is well-designed and thoroughly analysed, but I found it a bit difficult to follow. To do this research justice, I think that it should be published after the points below are addressed.

1. What’s the rationale for group vs individual antenatal care? The introduction includes a nice summary of the evidence available but does not mention the potential mechanism by which group care is hypothesised to improve gestational and perinatal outcomes.

2. While I appreciate that medical notes usually have many acronyms, people interested in reading this study may not necessarily know most of them, and reading a manuscript full of abbreviations can be quite time-consuming. Because PLOS One does not have a word count limit, I suggest keeping only a few universally known acronyms (e.g. RCT and WHO) and removing all the others from the text and figures.

3. Authors need to clarify the study aims from the beginning and to use descriptors consistently across the manuscript. For instance, the Methods section opens up with this sentence “To test the primary and secondary hypotheses [..]” although these hypotheses are described 5 pages later. I suggest adding a couple of sentences in the introduction to briefly mention the study hypotheses and related objectives, so that the reader has a clearer idea of the study aims. Also, are the primary hypothesis, primary outcome and primary analysis all linked somehow? If not, to avoid confusion I suggest using different adjectives, e.g. main analysis as opposed to ad-hoc analysis.

4. In the trial registration page (https://clinicaltrials.gov/ct2/show/NCT03154177) there is only one primary outcome, gestational age (“gestational length” in Table 4) and 8 secondary outcomes, whereas in the manuscript only 4 secondary outcomes are presented in tables 4 and 5. the following secondary outcomes are not presented in the main tables - apologies if I missed them somewhere else in the manuscript:

- Preterm 28-day and 42-day mortality rate

- Adherence to 6 week postnatal visit

- High-risk Women

- Newborns with neonatal morbidities

I suggest including all results in two tables, one with the findings of primary and secondary outcomes for the “group vs individual care” comparison and a very similar table for the “urine pregnancy testing vs ultrasound” comparison, while keeping the same outcome names using in the trial registration page.

5. This is a large-scale study and, understandably, the length of the Methods section reflects this. At the same time, I feel that some important details are missing. To preserve both readability and reproducibility, I suggest (i) moving most of the Methods section to a supplementary appendix, while keeping in the main text only the details that are essential to understand the Results and Discussion sections; and (ii) adding missing information to these supplementary Methods.

For example, the Sample Size section does not mention the following details

a) whether the trial was powered to detect an effect between group/individual care or between each of the 4 sturdy arms

b) what was the anticipated effect size relative to control arm(s), and from which previous studies was it obtained

c) the size of this inflation factor

d) the formula to calculate the final sample size from the inflation factor and the anticipated attrition constant, with a reference to the underlying methods paper

This trial has a large sample size in absolute terms, but if the anticipated effect size was very small it could still be underpowered, which could change the overall interpretation of the study. That is unlikely, but not impossible, and the reader currently does not have information to decide on this matter. While the most technical parts of this paragraph could be moved to the supplementary materials, an abridged version could stay in the main text, as it would be relevant to interpreting the study’s findings. An example PLOS One paper presenting the Methods in a more compact way and reporting the full protocol in the Appendix is available here: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0080561#s2

MINOR CONCERNS

This is a factorial cluster RCT, and the design could be mentioned in the title and the abstract.

Page 4. Ref #4 does not include the meta-analysis mentioned in the text. It looks like that would be reference #8 of this paper which includes the following sentence: “However, when the analysis was limited to the high quality studies (1 RCT and 1 observational study), African American women participating in group care had a significantly lower rate of preterm birth (2 studies (7, 8): pooled rates 8.0% vs. 11.1%, pooled RR 0.55; 95% CI 0.34–0.88).” consistently with what the authors report in the introduction of this paper. Please can authors replace reference #4 with https://pubmed.ncbi.nlm.nih.gov/27500348.

Page 7 Non-bipartite matching algorithm – to improve study reproducibility, please can you provide a specific reference (and/or the R package used for this, if applicable).

Page 10 Authors should briefly mention why a postnatal group visit was added, given that the primary purpose of the intervention was to improve gestational age.

Page 14. I agree with the choice to use generalized estimating equations model. To improve clarity, authors could explain in the method why the chose this method (I imagine to account for outcome correlation related to healthcare centres). An alternative could be using mixed effects models but this assumes a normal distribution of the data and this might not hold when using variables that are unlikely to be normally distributed, such as gestational age.

**********

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: Yes: Christopher J. Doig

Reviewer #2: No

Reviewer #3: 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.


4 Sep 2020

PONE-D-20-10387

Assessing the impact of group antenatal care on gestational length in Rwanda: a cluster-randomized trial

PLOS ONE

Response to Editor Comments

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

Thank you for the templates. We have reviewed the style requirements and made corrections accordingly.

2. You indicated that you had ethical approval for your study.

In your Methods section, please ensure you have also stated whether you obtained consent from parents or guardians of the minors included in the study or whether the research ethics committee or IRB specifically waived the need for their consent, or whether minors with children are legally allowed to consent for themselves.

Regarding consent for minors, we have updated the Ethics statement in the Methods section to describe the IRB-approved waiver of parental consent as follows: Ethical approval was granted by the Rwanda National Ethics Committee (No.0034/RNE/2017) and University of California, San Francisco Institutional Review Board (16-21177). Data were collected on women aged 15 and older presenting for ANC at the 36 study health care centers who provided written informed consent between May 25, 2017 and December 31, 2018. This study was permitted by the Rwanda National Ethics Committee to waive parental consent for pregnant minors ages 15 and older. No pregnant adolescents younger than 15 years were documented to have been invited to participate in the study.

3. In your Data Availability statement, you have not specified where the minimal data set underlying the results described in your manuscript can be found. PLOS defines a study's minimal data set as the underlying data used to reach the conclusions drawn in the manuscript and any additional data required to replicate the reported study findings in their entirety. All PLOS journals require that the minimal data set be made fully available. For more information about our data policy, please see http://journals.plos.org/plosone/s/data-availability.

Upon re-submitting your revised manuscript, please upload your study’s minimal underlying data set as either Supporting Information files or to a stable, public repository and include the relevant URLs, DOIs, or accession numbers within your revised cover letter. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories. Any potentially identifying patient information must be fully anonymized.

Regarding the data availability statement, we will make the minimal data set available on datadryad.org at https://doi.org/10.7272/Q67W69F1

4. Please note that in order to use the direct billing option the corresponding author must be affiliated with the chosen institute. Please either amend your manuscript or remove this option (via Edit Submission).

As a grantee of the Bill and Melinda Gates Foundation, it was our understanding that publishing fees would be paid directly by the Foundation. As such we have updated our record of this manuscript on the Chronos platform, the Foundation's portal to ensure open access publication and their payment of publication fees. If direct billing is not possible, please send an invoice to Elizabeth.butrick@ucsf.edu and we will submit it to the foundation.

5. We noted in your submission details that a portion of your manuscript may have been presented or published elsewhere:

'Table 1 was previously published in the Journal of Midwifery and Women's Health. Figure 1 was previously published in Gates Open Research. These publications are included as Related Manuscripts in this submission.'

Please clarify whether this publication was peer-reviewed and formally published.

If this work was previously peer-reviewed and published, in the cover letter please provide the reason that this work does not constitute dual publication and should be included in the current manuscript.

We appreciate the attention to detail; indeed, both Table 1 and Figure 1 have been previously published and peer reviewed. Because they do not include study data we thought it might be simpler to use the same information previously presented. Per your feedback, however, we have eliminated Table 1 from this manuscript, and inserted text with reference to the prior publication. Specifically, we state: During the second hour of the session, the provider and community health worker co-facilitated a discussion of health-related topics that aligned with the stage of pregnancy and the health issues of highest importance at that time. The full list of visit timing and topics at each visit are described elsewhere.13

Additionally, Figure 1 was previously published in our protocol paper. However, given some of the reviewers’ comments we have removed the figure as we believe it created confusion about the design. We have revised the text on page 6-7 to more accurately reflect the study design.

REVIEW COMMENTS TO THE AUTHOR

Reviewer #1: Thank you for the opportunity to read a well written paper on an important topic: addressing ANC and maternal/infant health in LMIC's/economically disadvantaged populations is very important. I do have some experience in healthcare in LMIC's in the African continent (but not Rwanda), cluster RCT's and GEE. I appreciate the complexity of the design, and the detailed explanations. I will focus more on technical issues related to design and analysis as this topic is not my clinical area of expertise.

1. This is a complicated design. I can not recall seeing an example of a cluster RCT with a factorial component. I remain unclear on aspects of your analysis. Usually a factorial design will report covariates for each factor, and an interaction estimate (between factors). The interaction effect is hopefully null, and then each factor effect can be reported independently. Was this done? Within a GEE, it is beyond my technical grasp how this would be setup given additional covariate for estimating random effects (intraclass correlations) as described in the manuscript. I think additional explanation is required, and perhaps supplemental analytical detail.

Thank you for highlighting this important design issue and giving us the opportunity to clarify the point. To clarify, the primary hypothesis of the study was not factorial, and the design was as a cluster randomized controlled trial. Conceptually, there is no reason to believe that ultrasound assessment would increase gestational age (the primary outcome) beyond the fact that early ultrasound assessment may rectify errors in LMP based GA. Further, there is no scientific evidence to further that line of argument. Additionally, ultrasound estimates of gestation were not included in the outcome variable. Thus, in line with our concept, and as stated in the protocol paper (https://gatesopenresearch.org/articles/3-1548) we have compared intervention (group antenatal care) vs. control (standard antenatal care).

We acknowledge the ambiguity in Figure 1 and in the main text about the design of the study. In the revised version, we have included clarifying text and removed the figure. The main intervention was group ANC and the study was powered to measure the effect of group care only, as hypothesized a-priori in the protocol paper. The ultrasound component of the intervention was primarily to support quality of care and improved GA estimation. Thus, in accordance to the trial design, the analysis compared group ANC care with standard care, generically that is intervention vs. control, respectively. Please refer to the changes on page 6-7 of the main text.

We chose not to implement ultrasound in all sites for two reasons. First, with limited resources and thirty-six sites it was not practical. Second, we hoped to use those sites with ultrasound to develop a GA estimation algorithm to refine GA estimates in sites without ultrasound, which is addressed in another manuscript under development. Because we could not introduce ultrasound in all sites, we assigned each pair to a similar sized pair and randomized half the matched pairs to receive ultrasound, so that the introduction would not disrupt the group care matching and to produce a balance of sites with ultrasound across groups.

2. The pair matching is an additional complicated aspect of the design. As I read your methods, it was unclear to me if the 4 factors for matching were considered separately, or were considered as a composite. I think the 'range' of values which were considered a 'match' should be reported (e.g. 50%+/-5% was considered a match for proportion of first ANC visits, monthly delivery volume was 60 +/-10, et cetera). If matching variables were treated independently (and not as a composite) I'm not certain of the validity of reporting variables used in matching in Table 2 (unless to confirm no failure in matching) as variables used for matching must be within the specified parameters of the original matching unless there was a technical failure. I believe this would apply to the first 4 variables in Table 2.

We acknowledge the complicated nature of this step in the methods and have revised the Randomization and masking section to more clearly describe the process as follows:

To optimize group comparability, 36 health centers were pair-matched. First, the study statistician applied a non-bipartite matching algorithm in R, using the nbpMatching package (available from: https://cran.r-project.org/web/packages/nbpMatching/index.html), including data on monthly ANC volume, monthly delivery volume, proportion of ANC visits initiated before 16 weeks gestation, and availability of screening tests as a composite. Thirty sites were matched with the non-bipartite matching algorithm [24] with the remaining 6 presented as multiple options, which were then reviewed by the study team that finalized matching based on ANC volume, rural/urban designation and distance to the nearest hospital. After matching was complete, pairs were further matched to quadruples. Then one site in each pair was assigned to group ANC and the other to standard ANC using random selection in R software 19. One pair of each quadruple was similarly randomly selected for implementation of ultrasound and urine pregnancy testing.

As the reviewer points out, Table 2 serves to confirm there was no failure in matching. Because the matching was an algorithm that accounted for a composite of the parameters of interest, we have not provided ranges for individual parameters as they are balanced across the composite. We offer Table 2 as an easier to digest metric for understanding the success of the pairing than the ranges suggested above. If the editors think it more appropriate, we could move Table 2 to Supplemental material.

3. There is a risk of selection bias given the loss to follow-up. It is reassuring that the relative proportions reported in Figure 2 are somewhat similar. Was there any analysis performed on characteristics between mothers who completed the study compared to those where mother-baby pair data was available?

We acknowledge the concern for potential selection bias and have performed chi-square tests to analyze the difference between women included in the final sample and those excluded due to lack or return for 2 ANC or missing delivery data. While some of the socio-demographic data pointed to statistically significant differences between the populations, we believe these differences are a result of the large population size and high number of degrees of freedom, rather than important clinically significant differences as the magnitude of difference is generally quite small. We have included this table at the end of this document (Appendix A), should the reviewer desire to examine these data more closely. We are willing to include this data in Supplemental material if the editor feels it would be important.

4. Why was GL/GA not calculated based on the estimate from the first ANC rather than based at the time of delivery? Given entry into the study based on this estimate, and issues of blinding (although assessors were unblinded, it is an objective measure of GL/GA estimate if using the estimate from first ANC and date of delivery).

We appreciate this comment as this is an issue our team has discussed at length. In analyzing our data, we compared gestational length as estimated by maternity providers at delivery compared to gestational length calculated using the date of delivery and gestational age estimated at ANC1. Further we have analyzed these data in sites with ultrasound to evaluate the best “predictors” of gestational age and have a manuscript under development on this topic. We found that gestational length as estimated by the maternity provider at birth constituted the most complete dataset with the highest proportion of biologically plausible values.

The reasons for this decision are threefold. One, maternity providers have the most information available to them. They have the EDD as estimated from the ANC1 gestational length, and they are able to assess the newborn for birthweight and development at the time of delivery. Two, gestational length estimation at ANC1 is primarily based on LMP, which women may or may not recall accurately. Three, we found that because performance-based incentives in the health system in Rwanda incentivize providers for women enrolled in ANC prior to 16 weeks gestation, there is pressure on the providers, which is passed on to the patients for women to come in early. This pressure may well result in women reporting in lower GAs than is actually accurate at the time of initiation of ANC. The misclassification of gestational age toward lower gestational ages at ANC1 would result in higher preterm birth rates.

Minor comment

5. In the results section, I think more technically accurate language would be that participant characteristics were statistically significant, but were not considered to be practically or clinically significant (I'm not sure of a technical meaning for insubstantial). I'm not a fan of statistical testing of demographic features between treatment arms, and your study might be a good example of permitting the reader to draw their own conclusion as to if any of these features are practically significant (rather than being drawn to a p value that perhaps better reflects size of the study).

Thank you for the recommendation. We agree and have thus removed the p-values from Table 3 (now Table 2). We have revised the text on page 17 to read: While various differences between the study groups' participant characteristics were statistically significant, we do not consider these differences to be practically or clinically significant (Table 2).

Reviewer #2: Abstract:

6. since this was a cluster randomised trial, the reported findings should also have the number of clusters included in the analysis. You should report this alongside the numbers of women in each arm, e.g. "4091 women in X clusters and 4752 women in Y clusters at control and intervention facilities, respectively, were included..."

Thank you. We have revised the first sentence of “Abstract/Findings” to read: 4091 women in 18 clusters and 4752 women in 18 clusters at control and intervention facilities had outcomes analyzed.

7. you need to specify the statistics represented by the "±1·6" and "±1·5" - is this the range of values? If so say so; if it is another statistic other than the range, please remove the ± designation.

Thank you for pointing this out. The statistic represented is the standard deviation. We have revised the sentence to read: Gestational length was equivalent in the intervention and control groups (39.3 weeks (SD 1.6) and 39.3 weeks (SD 1.5)).

Objectives:

8. the objectives make no mention of the ultrasound component of the intervention; indeed having read it, the description of the trial seems incomplete to me - it is a factorial cluster randomised trial testing the effect of group ANC with or without obstetric ultrasound at the health centre on gestational age at birth.

Please see Comment #1 for a detailed response to this issue. The ultrasound and community-based pregnancy testing, which were only expected to impact secondary outcomes, are mentioned in the last sentence of the Objectives section on page 5, and described in the Study Interventions section on page 10.

Methods:

9. in study endpoints where it is indicated that the study hypothesised that group ANC would increase mean GA at birth by 0·5±4·3 weeks, you should specify the statistic represented by "±4·3"; if it is not a range, then the ± designation should not be used.

Thank you for pointing this out. The statistic represented is the standard deviation. We have removed the ± designation throughout. We have revised the sentence, now in the Outcomes section on page 11 to read: …would increase mean gestational length by .5 weeks (with a standard deviation no larger than 4.3 weeks) compared to…

10. for the sample size calculation, you should indicate the cluster design-related parameters by which the sample size was inflated by to account for this design, with a justification for the choice.

At the design stage of this trial, we assumed an ICC of 0.01, based on the literature for conventional antenatal care versus group antenatal care. We assumed a loss-to-follow-up rate of 30-50%, working on the assumptions that 10-20% of all pregnancies end in miscarriage, and that 20-30% of women might deliver at another location. Given these assumptions, for 36 facilities and 222 deliveries per facility, the design effect that this study would have accounted for was 3.21. The study was powered to detect a 0.5 week (with a standard deviation of 4.3 weeks) difference in gestational length. For enhanced clarity, we have revised the sample size section of the manuscript, on page 13, copied below:

Sample Size:

At the design stage of this trial, we assumed an ICC of 0.01, and a possible effect size of half a week in gestational age based on a Cochrane review of conventional versus group antenatal care.[22] We assumed a loss-to-follow-up rate of 30%, based on the assumptions that 10-15% of all pregnancies end in miscarriage and 15-20% of women might deliver at another location. We calculated that for a two-tailed test, α=0·05, 1-β=80%, and a balanced 1:1 ratio of intervention (group antenatal care) to control (standard antenatal care) study participants, a minimum sample of n=1,163 eligible women per study group (intervention and control) was required. The sample size was increased to account for a cluster design effect of 3.21 (cluster randomization to study group by health center rather than by individual) and to account for a loss to follow-up rate of up to 50%, for a total required sample size per study group of 3,640.[23] As descriptive outcomes, the secondary outcomes were assessed in the available sample without hypothesis testing or multiple comparisons adjustment.

The ICC from actual data was much lower (0.00063) and with 246 deliveries per facility the actual design effect was 1.15. However, the observed difference was much lower (e.g., nearly null), 0.07 weeks, than the difference that this study was powered to detect.

11. please organise the sub-sections in the methods to follow the sequence recommended in the CONSORT guidelines

We have reordered the sub-sections in the methods section to more seamlessly follow the sequence recommended in the CONSORT guidelines. Thank you for the recommendation.

12. individual level bivariate analyses stratified by study group using Chi-square and Student’s t-test statistics are not appropriate for cluster-randomised trials because they do not adjust for clustering. In any case, statistical tests comparing groups in terms of baseline characteristics are not appropriate for randomised controlled trials.

Thank you for this comment. We agree that the lack of adjustment for clustering may be misleading. Another reviewer made a similar comment and thus we have opted to remove the p-values from Table 2. Please see Comment #5 for additional detail.

13. adjusting variables in a cRCT should be specified a priori and not based on post-hoc statistical considerations as described in the last paragraph of the methods.

We have not adjusted for any confounder in the final models in both the original and the revised version (Table 3). The two factors that we did adjust for relate to design aspects. The CRCT was pair-matched, hence the pairing had to be accounted for in the models. Further, the primary outcome was gestational length, which was measured for each birth, which in turn were clustered within facilities. Hence, we accounted for that clustering structure of the data in the statistical models. In other words, we only adjusted for “pairing” and “clustering.”

The adjustment that we mentioned in the last paragraph of the methods on page 18 of the original version, was part of the sensitivity analysis to demonstrate robustness of the results. We have modified the text to clarify this point, on page 15 of the revised manuscript:

“We developed a Directed Acyclic Graph to identify potential confounders and mediators for which data were available. In descriptive analyses, we investigated the factors on which the intervention and control arms were statistically significantly different. The final model was adjusted for pairing and clustering only. In sensitivity analyses, we additionally adjusted for potential confounders or mediators to demonstrate that the results were unaffected by residual confounding that may have resulted due to unbalanced study arms.”

In addition, we would like to note that given the null results, it is unlikely that any single or group of suppressor variables would have explained the observed results.

Results:

14. the CONSORT diagram is not appropriate for a cluster RCT; it should include the numbers of clusters at each stage and the mean or median number of women per cluster with the SD or IQR.

Thank you for drawing our attention to this important omission. The CONSORT diagram has been revised accordingly.

15. the statistical tests comparing groups in table 3 are not appropriate in a cRCT

Another reviewer made a similar comment and thus we have removed the p-values from Table 2. Please see comment #5 for additional detail.

16. avoid the ± designation all through the results and specify what the reported statistic is, whether range, SD or SE. For continuous outcomes, the mean and SD should be reported in the descriptive tables, and the mean and SE for inference (e.g. in Table 4, Table 5, Table 6, and the results of primary and secondary outcomes).

Thank you for this comment. For all descriptive estimates such as means, we have removed the ± designation throughout and reported SDs. For inferential estimates (i.e. model based results) we reported the confidence intervals in Tables 3-4.

17. see comment above about adjusting variables in randomised controlled trials.

Please see response above to comment # 13.

Reviewer #3: This is an important paper summarising the results of a large cluster randomised controlled trial (RCT) assessing the effects of group antenatal care among women in Rwanda on child gestational length. The study is well-designed and thoroughly analysed, but I found it a bit difficult to follow. To do this research justice, I think that it should be published after the points below are addressed.

18. What’s the rationale for group vs individual antenatal care? The introduction includes a nice summary of the evidence available but does not mention the potential mechanism by which group care is hypothesised to improve gestational and perinatal outcomes.

Thank you for this comment. We have added the rationale in the introduction at the end of the first paragraph, appending the following sentence:

Group ANC is hypothesized to positively impact preterm birth rates and other outcomes among women at elevated psychosocial risk due to 3 main features of the model: 1) greater social support between women who are linked via the group; 2) more total ANC-associated time spent in educational activities in facilitated group discussions; and 3) attention to key elements of person-centered care, including respect and safety, empowerment, and participation.5–7 These elements create a more positive pregnancy care experience which may encourage ANC attendance and thus create additional opportunity for risk assessment by providers.

19. While I appreciate that medical notes usually have many acronyms, people interested in reading this study may not necessarily know most of them, and reading a manuscript full of abbreviations can be quite time-consuming. Because PLOS One does not have a word count limit, I suggest keeping only a few universally known acronyms (e.g. RCT and WHO) and removing all the others from the text and figures.

We have spelled out the following acronyms in the text ANC, PNC, PTBi, LMIC, MOH, UPT, PTB, LBW, SGA, REDCap, GA, and GEE. We retained the acronyms: RCT, WHO, SPSS, STATA, and HIV, and “R” software

We also changed “gestational age at birth” to “gestational length” throughout text to be consistent with article title.

20. Authors need to clarify the study aims from the beginning and to use descriptors consistently across the manuscript. For instance, the Methods section opens up with this sentence “To test the primary and secondary hypotheses [..]” although these hypotheses are described 5 pages later. I suggest adding a couple of sentences in the introduction to briefly mention the study hypotheses and related objectives, so that the reader has a clearer idea of the study aims. Also, are the primary hypothesis, primary outcome and primary analysis all linked somehow? If not, to avoid confusion I suggest using different adjectives, e.g. main analysis as opposed to ad-hoc analysis.

Thank you. We have made a number of revisions to improve clarity and consistency related to our study aims.

As per the study's a priori hypotheses, this paper reports on primary and secondary outcomes of our CRCT only. The primary analysis presents gestational length, the study primary outcome, on which the study was powered. The intervention tested is the effect of group antenatal care on gestational length at birth. We also analyzed secondary outcomes, as outlined in our protocol paper including the effect of group antenatal care on rates of preterm birth, mortality among preterm neonates, attendance at 4 ANC visits, initiation of ANC before 14 weeks, attendance at 6-week postnatal care visits, identification of women at high risk during pregnancy, and cesarean section. We are unable to report on newborn morbidities as planned due to data quality issues.

In addition to the intervention of interest, group antenatal care, both ultrasound examination by health center providers and community-based urine pregnancy testing were implemented in half of the sites in each study group. Additional secondary analyses were conducted to see if group antenatal care, ultrasound examination and/or community-based urine pregnancy testing were associated with the secondary outcomes of attendance at 4 ANC visits and initiation of ANC before 14 weeks.

We have made modifications to the paper to help improve clarity including:

• In the “objectives” subsection of the Introduction: The sentence “Intrigued by lower rates of preterm birth among high-risk American women who participated in group antenatal care, the Preterm Birth Initiative-Rwanda aimed to test the primary [added ‘primary’] hypothesis that Rwandan women receiving antenatal care at health centers that offer group antenatal care would experience increased gestational length compared to women receiving antenatal care at health centers that provide the standard, individual model of care. Rwandan stakeholders preferred to offer all women at each health center the same model of care.”

• Added to this paragraph, on page 5-6: We also explored whether group antenatal care affected the secondary outcomes of preterm birth, mortality among preterm neonates, attendance at 4 antenatal care visits, attendance at the first antenatal care visit before 14 completed weeks gestation, attendance at a six-week postnatal care visit at health facilities, identification of women as at high risk at any antenatal care visit, and caesarean section rates among enrolled women. We intended to examine newborn morbidities but our data sources were not adequate to do so. Further, at half of the health centers included in this study we implemented community-based urine pregnancy test by community health workers and basic obstetric ultrasound by nurses and midwives and conducted secondary analyses to see if these interventions affected the secondary outcomes of attendance at 4 antenatal care visits and initiation of antenatal care before 14 completed weeks.

21. In the trial registration page (https://clinicaltrials.gov/ct2/show/NCT03154177) there is only one primary outcome, gestational age (“gestational length” in Table 4) and 8 secondary outcomes, whereas in the manuscript only 4 secondary outcomes are presented in tables 4 and 5. the following secondary outcomes are not presented in the main tables - apologies if I missed them somewhere else in the manuscript:

- Preterm 28-day and 42-day mortality rate

- Adherence to 6 week postnatal visit

- High-risk Women

- Newborns with neonatal morbidities

I suggest including all results in two tables, one with the findings of primary and secondary outcomes for the “group vs individual care” comparison and a very similar table for the “urine pregnancy testing vs ultrasound” comparison, while keeping the same outcome names using in the trial registration page.

Thank you for your careful review of both our paper and our protocol. We have revised the tables according to the reviewer’s suggestions, as we agree it is important to report on the primary and all the listed secondary outcomes.

The data for the outcomes preterm mortality and morbidity is too scanty to reliably make any inferences, and we have noted as much in the revised manuscript.

However, we have reported figures on newborn mortality, though the majority of the available data was at birth, rather than at 28 and 42 days as planned, due to insufficient follow-up data. We were unable to systematically collect newborn morbidity data and thus have left it out of the table, and inserted an explanation on page 12 of the omission and addressed it in the discussion.

These data limitations were due, at least in part, to the fact that one of the expected data sources was the national SMS reporting of newborn morbidities and mortalities by community health workers. Unlike all of our other data sources, where study data collectors could access paper copies of facility registers and patient files without impeding the work of health center staff, retrieving SMS data required significant time and cooperation from facility-based data managers. Our data collection team was unable to effectively gain the systematic cooperation of these data managers, and thus in our data system not a single newborn morbidity report appeared. Further, reports from our field staff reveal that the Rapid SMS system was often down in rural areas, limiting their ability to access this data. Rather than report this patently deficient data point, we opted to omit it from the table, and have instead acknowledged the limitation in the text on page 26.

As suggested, we have included all the secondary outcomes in the revised main outcomes table. The outcomes table for the analysis by ultrasound and pregnancy testing versus no ultrasound or pregnancy testing contains only the secondary outcomes of gestational age at entry to care and proportion of women attending at least 4 antenatal visits, as these are the only outcomes these were expected to be associated with.

22. This is a large-scale study and, understandably, the length of the Methods section reflects this. At the same time, I feel that some important details are missing. To preserve both readability and reproducibility, I suggest (i) moving most of the Methods section to a supplementary appendix, while keeping in the main text only the details that are essential to understand the Results and Discussion sections; and (ii) adding missing information to these supplementary Methods.

For example, the Sample Size section does not mention the following details

a) whether the trial was powered to detect an effect between group/individual care or between each of the 4 sturdy arms

b) what was the anticipated effect size relative to control arm(s), and from which previous studies was it obtained

c) the size of this inflation factor

d) the formula to calculate the final sample size from the inflation factor and the anticipated attrition constant, with a reference to the underlying methods paper

This trial has a large sample size in absolute terms, but if the anticipated effect size was very small it could still be underpowered, which could change the overall interpretation of the study. That is unlikely, but not impossible, and the reader currently does not have information to decide on this matter. While the most technical parts of this paragraph could be moved to the supplementary materials, an abridged version could stay in the main text, as it would be relevant to interpreting the study’s findings. An example PLOS One paper presenting the Methods in a more compact way and reporting the full protocol in the Appendix is available here: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0080561#s2

Thank you. These points have been addressed in Comment #10 in response to another reviewer.

Further, we appreciate the comment on the power. The design effect for the ICC of the study sample (0.00063 and 8843 respectively) for 36 clusters (246 women per cluster) is 1.15435. We originally assumed an ICC of 0.001, so, although we had a smaller ICC, we also had a much smaller effect than the 0.5 week effect we designed for. When designing the trial we chose the 0.5 week effect size as the smallest clinically relevant effect we thought might be attainable.

MINOR CONCERNS

23.This is a factorial cluster RCT, and the design could be mentioned in the title and the abstract.

Please see comment #1. We appreciate this perspective, but have retained the title as the study was designed as a CRCT.

24. Page 4. Ref #4 does not include the meta-analysis mentioned in the text. It looks like that would be reference #8 of this paper which includes the following sentence: “However, when the analysis was limited to the high quality studies (1 RCT and 1 observational study), African American women participating in group care had a significantly lower rate of preterm birth (2 studies (7, 8): pooled rates 8.0% vs. 11.1%, pooled RR 0.55; 95% CI 0.34–0.88).” consistently with what the authors report in the introduction of this paper. Please can authors replace reference #4 with https://pubmed.ncbi.nlm.nih.gov/27500348.

Thank you for your attention to detail. This has been corrected in the revised manuscript.

25. Page 7 Non-bipartite matching algorithm – to improve study reproducibility, please can you provide a specific reference (and/or the R package used for this, if applicable).

Thank you. Please see Comment #2 for additional discussion on this topic. Our revised text there includes a reference to Lu et al 2011, (Optimal Nonbipartite, Matching and Its Statistical Applications, The American Statistician, 65:1, 21-30). We have also included a link to the nbpMatching package:https://cran.r-project.org/web/packages/nbpMatching/index.html

26 Page 10 Authors should briefly mention why a postnatal group visit was added, given that the primary purpose of the intervention was to improve gestational age.

Thank you for this insightful comment. We have revised a sentence in second paragraph of Group Antenatal Care and Postnatal Care subsection to include this explanation:

The Technical Working Group hoped that the social support fostered among women in the same antenatal group would continue into the postnatal period and motivate women to seek care; for this reason, a postnatal group visit was included in the model even though a postnatal visit was not expected to impact the primary outcome (gestational length).

27. Page 14. I agree with the choice to use generalized estimating equations model. To improve clarity, authors could explain in the method why the chose this method (I imagine to account for outcome correlation related to healthcare centres). An alternative could be using mixed effects models but this assumes a normal distribution of the data and this might not hold when using variables that are unlikely to be normally distributed, such as gestational age.

We appreciate the concurrence of the reviewer with our analytical approach. The robust GEE and the mixed effect models are the two suitable options to analyze this data. We chose the former because we wanted to generate robust estimates that controlled for design effect. We chose the GEE to generate one overall intervention effect across all pairs of facilities. In addition, to the point the reviewer highlighted, we would like to state that our interest was not to generate intervention effect separately for each pair of facilities, or how this effect might have differed between pairs, in which case a mixed effect model with a random intercept/slope for each pair would have been relevant.

Submitted filename: Response to Reviewers28Aug2020_PONE-D-20-10387.docx

3 Nov 2020

PONE-D-20-10387R1

Assessing the impact of group antenatal care on gestational length in Rwanda: a cluster-randomized trial

PLOS ONE

Dear Dr. Lundeen,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Dec 18 2020 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

  • 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: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Seth Adu-Afarwuah

Academic Editor

PLOS ONE

[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: (No Response)

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 #1: Yes

Reviewer #2: Yes

**********

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

Reviewer #1: Yes

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 #1: Yes

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: Yes

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: Thank you for your thoughtful comments. I've had the opportunity to review responses to my comments, and the comments of other reviewers. I'd like to congratulate you on conducting an important study with an elegant design.

Reviewer #2: Abstract: correct the typo in the first line of 'Findings'; I suggest it reads: "A total of 4091 women in 18 control clusters and 4752 women in 18 intervention clusters were included in the analysis..."

Background line 4: "an individually randomized" instead of "individual randomized"

Objectives: a lot of the explanations now included in the objectives, although very useful, do not belong there. For example, the following sentence should be included in the justification for the cluster-randomised design, such as in the second sentence of the 'trial design' section (words in brackets are my own suggestion): "(We opted for a clustered design because) Rwandan stakeholders preferred to offer all women at each health center the same model of care." The following sentence should also be moved to an appropriate place in the methods: "Further, at half of the health centers included in this study we implemented community-based urine pregnancy test by community health workers and basic obstetric ultrasound by nurses and midwives and conducted secondary analyses to see if these interventions affected the secondary outcomes of attendance at four antenatal care visits and initiation of antenatal care before 14 completed weeks."

Methods: description of the sample size calculation is still inadequate. This was a cluster-randomised trial; the description should include at least two of the following: the number of clusters per arm; the number of observations per cluster; the number of individuals per arm. At the moment, only the number of individuals per arm is included, giving the reader no idea of the number of clusters per arm or observations per cluster. Please look at the reporting in other examples of studies of the same design e.g. 10.1371/journal.pmed.1001018.

According to your response to peer review, you are no longer doing this, therefore this should be removed from your 'statistical analysis' section: "We conducted individual level bivariate analyses stratified by study group using Chi-square and Student’s t-test statistics for categorical and continuous data, respectively, to assess study group comparability. Similar unadjusted bivariate analyses were conducted for the primary and secondary outcomes." - please go through the whole manuscript carefully to ensure that the changes you have made are reflected in the latest version and are consistent all through.

Results: for the description of the sample in the first paragraph of the results, please report the mean and range or median and IQR per cluster, not mean and SD - the SD is not very meaningful in this context (fine to report SD elsewhere). Your CONSORT diagram is still incomplete - it should include the number of clusters with mean+range or median+IQR of women per cluster all through; see the CONSORT diagram in 10.1371/journal.pmed.1001018 for example.

When you report the continuous outcomes, you should use standard errors (SE) instead of standard deviations (SD); SDs are for description, SEs are for inference (and you are making inference on the outcomes).

**********

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: Christopher James Doig, Professor, Departments of Critical Care Medicine, and Community Health Sciences, University of Calgary

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 Nov 2020

PONE-D-20-10387R1

Assessing the impact of group antenatal care on gestational length in Rwanda: a cluster-randomized trial

PLOS ONE

RESPONSE TO REVIEWERS

Reviewer #1: Thank you for your thoughtful comments. I've had the opportunity to review responses to my comments, and the comments of other reviewers. I'd like to congratulate you on conducting an important study with an elegant design.

Reviewer #2: Abstract: correct the typo in the first line of 'Findings'; I suggest it reads: "A total of 4091 women in 18 control clusters and 4752 women in 18 intervention clusters were included in the analysis..."

Thank you for your careful attention to detail. We have corrected the typo and appreciate the suggestion.

Background line 4: "an individually randomized" instead of "individual randomized"

Corrected.

Objectives: a lot of the explanations now included in the objectives, although very useful, do not belong there. For example, the following sentence should be included in the justification for the cluster-randomised design, such as in the second sentence of the 'trial design' section (words in brackets are my own suggestion): "(We opted for a clustered design because) Rwandan stakeholders preferred to offer all women at each health center the same model of care." The following sentence should also be moved to an appropriate place in the methods: "Further, at half of the health centers included in this study we implemented community-based urine pregnancy test by community health workers and basic obstetric ultrasound by nurses and midwives and conducted secondary analyses to see if these interventions affected the secondary outcomes of attendance at four antenatal care visits and initiation of antenatal care before 14 completed weeks."

Thank you for the suggestions. In rereading we understand that some of the additions we put in in response to previous reviewer comments may have overcompensated. We further appreciate the specificity of this reviewer’s suggestions and have made the following changes:

As suggested, we moved the explanation of Rwandan stakeholders’ preference for uniform care at any health center to the trial design justification, on page 6. We have also edited this section to address only the objectives and moved the information about ultrasound and urine pregnancy section to the methods, describing in trial design the distribution of these interventions in a balanced manner on page 6-7.

Methods: description of the sample size calculation is still inadequate. This was a cluster-randomised trial; the description should include at least two of the following: the number of clusters per arm; the number of observations per cluster; the number of individuals per arm. At the moment, only the number of individuals per arm is included, giving the reader no idea of the number of clusters per arm or observations per cluster. Please look at the reporting in other examples of studies of the same design e.g. 10.1371/journal.pmed.1001018.

We have revised the sample size section on page 13 to reflect the underlying assumption of 36 health centers total, or 18 per study group, and an average of 202 women per cluster.

According to your response to peer review, you are no longer doing this, therefore this should be removed from your 'statistical analysis' section: "We conducted individual level bivariate analyses stratified by study group using Chi-square and Student’s t-test statistics for categorical and continuous data, respectively, to assess study group comparability. Similar unadjusted bivariate analyses were conducted for the primary and secondary outcomes." - please go through the whole manuscript carefully to ensure that the changes you have made are reflected in the latest version and are consistent all through.

Thank you. We apologize for the confusion. We did conduct bivariate analysis for comparability, but removed the Chi-square and t-tests, as well as all unadjusted results for outcomes, and have now thoroughly reviewed the manuscript for these inconsistencies and removed references to results that are not presented in the paper.

Results: for the description of the sample in the first paragraph of the results, please report the mean and range or median and IQR per cluster, not mean and SD - the SD is not very meaningful in this context (fine to report SD elsewhere). Your CONSORT diagram is still incomplete - it should include the number of clusters with mean+range or median+IQR of women per cluster all through; see the CONSORT diagram in 10.1371/journal.pmed.1001018 for example.

Thank you. We appreciate the clarifications from this reviewer and apologize for the confusion. We have replaced SDs with ranges in the first paragraph of the results section, and updated the CONSORT throughout to include cluster size and ranges at each level.

When you report the continuous outcomes, you should use standard errors (SE) instead of standard deviations (SD); SDs are for description, SEs are for inference (and you are making inference on the outcomes).

Thank you for the clarification. We have updated the continuous outcomes in Tables 3, S1, and S3 to provide SEs rather than SDs for continuous outcomes as well as more clearly labelling the column headers for the reader.

We have similarly cleaned up the column headers in Tables 1 and 4 for clarity.

Submitted filename: Response to Reviewers 2.docx

20 Jan 2021

Assessing the impact of group antenatal care on gestational length in Rwanda: a cluster-randomized trial

PONE-D-20-10387R2

Dear Dr. Lundeen,

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,

Seth Adu-Afarwuah

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: (No Response)

**********

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

Reviewer #2: (No Response)

**********

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: (No Response)

**********

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: (No Response)

**********

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


22 Jan 2021

PONE-D-20-10387R2

Assessing the impact of group antenatal care on gestational length in Rwanda: a cluster-randomized trial

Dear Dr. Lundeen:

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. Seth Adu-Afarwuah

Academic Editor

PLOS ONE

This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

https://www.researchpad.co/tools/openurl?pubtype=article&doi=10.1371/journal.pone.0246442&title=Assessing the impact of group antenatal care on gestational length in Rwanda: A cluster-randomized trial&author=&keyword=&subject=Research Article,Medicine and Health Sciences,Women's Health,Maternal Health,Antenatal Care,Medicine and Health Sciences,Women's Health,Maternal Health,Pregnancy,Medicine and Health Sciences,Women's Health,Obstetrics and Gynecology,Pregnancy,Medicine and Health Sciences,Women's Health,Maternal Health,Birth,Preterm Birth,Medicine and Health Sciences,Women's Health,Obstetrics and Gynecology,Birth,Preterm Birth,Medicine and Health Sciences,Women's Health,Maternal Health,Pregnancy,Pregnancy Complications,Preterm Birth,Medicine and Health Sciences,Women's Health,Obstetrics and Gynecology,Pregnancy,Pregnancy Complications,Preterm Birth,Biology and Life Sciences,Physiology,Physiological Parameters,Body Weight,Birth Weight,Medicine and Health Sciences,Women's Health,Maternal Health,Birth,Medicine and Health Sciences,Women's Health,Obstetrics and Gynecology,Birth,People and Places,Geographical Locations,Africa,Rwanda,People and Places,Population Groupings,Age Groups,Children,Infants,People and Places,Population Groupings,Families,Children,Infants,Medicine and Health Sciences,Epidemiology,Medical Risk Factors,