The RCT was conducted at the obstetric service of the Hospital Universitario San Ignacio, a tertiary care referral centre in Bogota, Colombia. Women attending the admission unit of the obstetrics' clinic for delivery were invited to participate.
In accordance with the protocol, the examining intern or resident would invite all eligible women to participate in the study. Recruitment rates varied greatly among different interns and residents. Because of the frequent rotation of interns and residents through the admission unit, interns and residents attended standardised monthly training sessions on how to enrol patients on the study.
Inclusion criteria included: absence of life threatening events at admission interview (such as placental abruptio, prolapsed cord, or eclampsia); gestational age ≥36 weeks based on the best available estimation such as a reliable last menstrual period date or an appropriate ultrasound; projected permanence in Bogota during the month following delivery; and willingness to participate expressed in a written informed consent. There was no age restriction for participants. However, for women under 18 years of age, an adult witness – most frequently a next of kin, was also asked to participate in the informed consent process.
Exclusion criteria included: a clinical diagnosis of any systemic or gynaecological bacterial infection; use of systemic antibiotics during the week prior to admission; rupture of amniotic membranes or uncertainty of their integrity; or a cervical dilatation >7 cm.
The trial protocol was approved by the Institutional Review Board of the Medical School of the Javeriana University and by the staff of the Clinical Epidemiology and Biostatistics Unit.
Women willing to participate were offered a range of additional services during follow-up. These services were designed to improve outcome detection and adherence to the study. All women received a booklet that addressed questions frequently raised by women during labour and puerperium (30 days following delivery), and informed on early signs for maternal and neonatal infections. The development of the booklet involved semi-structured interviews carried out by a trained nurse who would elicit the issues that worried women during the first day, first week and first month after delivery. The team that designed the booklet included graphic designers and a social communications specialist. The booklet was tested and printed prior to the beginning of the RCT. Participants were offered two programmed health care visits where their concerns could be discussed with a professional nurse. During the visits, the nurse examined the mother and the baby and assessed outcomes. Systematic telephone reminders for these visits were scheduled before the 1st and 4th week after delivery. Women received instructions on how to access a 24-hour paging service which allowed them to contact a health care provider to address any concerns, seek support, inform about problems, or get advice on issues such as emergency medical attention. This allowed retrieving relevant information of visits to other healthcare providers and contacting them to collect data. The local branch of La Leche League International, a volunteer organization delivering support and education on breastfeeding, offered regular free educational sessions for participants. All participants were offered subsidised screening for neonatal hypothyroidism; when this RCT was done, screening for thyroid disease was not mandatory nor covered by health maintenance organisations in Colombia.
Assignment and follow-up procedures
Randomisation was done in blocks of 2 (20%), 4 (60%) and 6 (20%) using Ralloc® allocation software. Once the participant had completed the informed consent process, an opaque envelope with sequential numbering and instructions was opened. Women randomised to the intervention group received a 1 litre Travad® 2.5% sodium chloride solution enemas applied by a nurse assistant prior to been taken to the labour wards, where participants in both groups would thereafter receive the same care.
During the first and third week after delivery, reminders to attend scheduled appointments at the outpatient primary care clinic were delivered by phone and mail. We used a standardised telephone survey to assess if participants had used any other health services or been diagnosed with any particular condition. Newborns were screened for hypothyroidism during the first scheduled visit when the mother agreed to it. Results for the screening tests were scheduled to be delivered at the second visit, but these were also couriered with an explanatory note if the appointment was missed. The screening scheme covered confirmation diagnostic tests, when necessary.
The primary outcomes were the diagnosis of infections in newborns or women during the month following delivery. A neonatal outcome was positive if during the first month of life the child was prescribed systemic antibiotics. A neonatal outcome was also positive when the child was diagnosed with any of the following clinical conditions: ocular infection (purulent drainage in the eye after the sixth day of delivery), umbilical infection (foul smell with periumbilical erythema), skin infection (cellulitis or impetigo), lower or upper respiratory tract infection, intestinal infection, meningitis or sepsis.
A puerperal outcome was positive when, during the first month after delivery, a health care provider diagnosed the women with any of the following: dehiscence of the episiorraphy suture, purulent effusion from the episiorraphy, urinary tract infection, pelvic inflammatory disease, or vulvovaginitis.
The primary outcome of the study was an aggregated maternal and neonatal infection rate: either the mother or the newborn had an infectious outcome (combined infection rate) .
A team member visited participating women and newborns in hospital on a daily basis. Throughout the trial, trained research assistants using standardised questionnaires, registered data from telephone calls, hospitalisations, follow-up visits, and any communications with the participants, their families, or their health care providers.
Masking the use of enemas was unfeasible. However, we made efforts to conceal the intervention by not separating documents with information on the allocation from those outcome data collection, by training the team's supporting clinical team (professional nurse, family medicine residents, family medicine staff, and consulting dermatologist) to avoid enquiring in ways that would unmask the allocation. Health care providers in other settings, such as physicians at emergency wards, paediatricians and medics at outpatient clinics were unaware of the allocation and frequently of the specific objectives of the study. Except for the intervention, participants received the same health care, and data retrievers would remain unaware of individual allocations. Interventions remained coded for the analysis and the code was broken once the analysis was completed. Input between data at recruitment and allocation was done weeks before the collection of data on outcomes at follow up.
Sample size determination
We were unable to find reliable data on the incidence of baseline infection rates for puerperal women or newborns so we did a pilot study to have base data that would allow a good estimate of frequencies for sample size calculations. The pilot study, which included the first 44 participants of the control group, estimated the combined infection rate of puerperal women and newborns at 46 percent . Using 5% significance and a 80% power, a sample of 394 participants distributed in two parallel groups was estimated to be required to detect a relative difference of 25% in combined infection rates of women and newborns, following the formulas provided by Duppont and Plummer. Assuming 4% of the participants would be lost to follow-up, an estimated total sample of 410 women was required.
Data management and analysis
The database created in Epi-Info v 6.04 b was fed using double data entry and transferred to Stata 5.0© using Stata Transfer 4.0©. A Shapiro-Wilk test was used to determine if the distribution of continuous variables was normal. Non-normal distributions were transformed using a log transformation, and if the distribution was persistently non-normal, a Mann-Whitney test was used to compare groups. Bivariate analyses were done using the chi-square test or Fisher's exact test. Power calculations were done using specialized software developed at the Clinical Epidemiology Unit at the Javeriana University.