Study design and population
The details of the PROFEG trial have been described elsewhere22
and only the main features are given here. The trial was a pragmatic RCT to compare two iron administration policies (routine iron prophylaxis vs screening and treatment for anaemia during pregnancy) on MCH in Maputo, Mozambique. The trial was carried out in two health centres, 1o de Maio in Maputo City, the capital (November 2006—October 2008), and Machava 2 in Maputo Province (June 2007—October 2008), Mozambique. The completion of collection of birth data continued until 2012. The health centre of Machava 2 in Maputo province is close to Maputo city. The population is urban and semiurban, and malaria is endemic in both areas. A seasonal increase of malaria is usually observed towards the end of the rainy season (February–April).23
In the study area, all woman were eligible to attend prenatal care. The usual care recommendations at the time of the trial included daily prophylactic iron–folate supplementation (60 mg+400 μg) throughout pregnancy; one dose of mebendazol 500 mg for intestinal parasite; three doses of sulfadoxine pyrimethamine for malaria prophylaxis (started around 20 weeks gestation, or when quickening occurs or when the fetal heart is heard); haemoglobin measurement (Lovibond is routinely used) and syphilis screening at the first prenatal visit and three doses of tetanus vaccine (at the fifth and seventh months and at delivery). If malaria was suspected during prenatal consultations, it was diagnosed by laboratory tests and clinical signs. In most health centres, including our study centres, HIV testing was offered.22
Antiretroviral (ARV) drugs were provided by various international organisations, but we do not have information of how many women received treatment during pregnancy. The recommendation was to give ARV (Nevirapine) at delivery to prevent mother–child transmission.
Recruitment of study participants
Pregnant women attending their first prenatal visit were the target group. During the routine early morning health education sessions, all women who came for their first prenatal visit were given general information about the study. Recruitment into the study occurred during individual consultations and was carried out by study nurses who were employed and trained by the project. In the 1° de Maio health centre, the women visited the study nurses after their routine prenatal care consultations with the MCH nurses. In Machava, the study nurse and the routine MCH nurse saw the women in the same room. The study nurses checked for the women's eligibility to participate in the study. The exclusion criteria were: women with high obstetric risk and those aged less than 18 years. If eligible, the nurses asked the women to join the study. Oral and written informed consent was obtained. Three types of women were missed from the study: women whom MCH nurses sent back home because of too early pregnancy, women who did not go to the study nurse and women who refused the study.
The women were randomised into either the Routine iron group (ie, routine iron prophylaxis from the first to the last prenatal visit) or the Selective iron group (ie, regular screening for haemoglobin level and treatment for anaemia). Researcher (OA) used the STATA statistical software (StataCorp LP, Texas, USA) to generate sequential random numbers separately for the two centres, and the women were assigned to either of the groups with a probability of 50%. The codes for the groups were put into sealed and numbered opaque envelopes; the number was the woman’s study number and was repeated in the documents in the envelope. The envelope contained a study identification card (yellow for the Routine iron group and pink for the Selective iron group, 10×20 cm) and the informed consent form.
We did not have up-to-date reliable baseline data of pregnant women's and newborns’ health in Maputo or of the effects of iron on pregnancy and birth outcomes. Thus, we used different estimates of the baseline values for preterm delivery, low birth weight, clinical malaria and perinatal mortality to calculate the sample size, with power (85% and 90%), significance level of 5% and the size of the difference to be detected (20% and 30%). Based on these calculations and the expected feasibility, we decided on a sample size of 2000 women in each group to be enough to measure clinically meaningful effects. The STATA statistical software was used to estimate the sample size. A table showing the various baseline assumptions used for power calculation and in estimating the sample size for the study is included as online supplementary appendix 1.
On each prenatal visit, women in the Routine iron group received 30 tablets (supply of 1 month) of 60 mg ferrous sulfate plus 400 μg of folic acid per day combined in one tablet. In the Selective iron group, women's haemoglobin levels were measured at each visit by the study nurses using a rapid haemoglobin measure, HemoCue Hb 201+, (Hemocue AB, Ängelholm, Sweden). If the haemoglobin was 9 g/dl or more, they received 30 tablets of 1 mg of folic acid per day. If their haemoglobin was below the cut-off of <9 g/dl haemoglobin, they received a monthly double dose of iron (60 mg+60 mg) for the treatment of anaemia. Folic acid 1 mg tablets were used because at the time of the trial pure folic acid was not licensed in Mozambique in 400 μg tablets. The tablets were given in a plastic bag having the drug’s name and dosage on it.
Data collection and follow-up
Data were collected on standard study data forms by three methods: (1) study nurses abstracted prenatal data from mothers’ maternity cards, (2) study nurses asked women additional questions at the time of the prenatal visits and (3) study nurses or researchers collected birth data afterwards from hospital birth records. Delivery nurses were informed of the study and asked to put the delivery cards into a separate study box. The study women were to be identified by the colour of the identification card stapled to their maternity card. However, this did not succeed very well. By excluding estimated late miscarriages (5%), early stillbirths (3%) and home births (10%), we should have received delivery data for 3547 women (82%) of the 4326 women who participated in the trial. We received birth data for only 2258 (64% of the estimated 3547) women.
The primary outcomes were preterm delivery (delivery <37 weeks of gestation and low birth weight (<2500 g); data on weight came from the birth records; for gestation weeks, various routine data sources were used (see below). Originally, we had malaria activation as a primary outcome, but the pilot showed that it was not feasible. Secondary outcomes were perinatal mortality (as available from our data collection forms; unlikely to cover early stillbirths or neonatal deaths occurring at home); complications during pregnancy and labour; symptoms suggestive of malaria (fever, headache, cold/chills, nausea/vomiting and body aches) and self-reported malaria during pregnancy (the woman was asked for diagnosed malaria since her last visit).
In the prenatal visits, routine MCH nurses determined gestational weeks in various ways, even though all ways were not systematically noted down. In the first prenatal visits, the date of the last menstrual period, uterine fundal height, assumed date of delivery and length of gestation (best estimate) were noted. The study nurses abstracted all this information and the best estimate was used in this paper. In birth records, the last menstrual period, date of fertilisation, assumed date of fertilisation and length of gestation were to be given by the delivery nurses. However, these data were very poorly filled and only 681 (30%) of the women with delivery data had their gestational weeks recorded at birth. Thus, the gestational weeks for women without that information were estimated from dates using the following algorithm: gestational weeks at first visit in days+days between the first visit and delivery; the days were then transformed into weeks. For some women (n=196), the date of delivery was not available. In these cases, the date of discharge from the hospital after delivery (minus the length of stay at the hospital; n=22) or the date of admission to the hospital (n=60 women who did not have the date of discharge) was used.
The women were instructed and encouraged at each visit to take the tablets they were given. Women allocated to the Routine iron group could refuse to take the iron tablets; in that case, they were classified as non-compliant with the intervention. Women who belonged in the Selective iron group and who wanted iron (even if their haemoglobin level was not below the cut-off level) were given iron; they were classified as non-compliant with the intervention. The following questions were asked on each visit: ‘Was hemoglobin measured?’; ‘Was iron/folic acid given to the woman?’; ‘Number of iron/folic acid tablets given?’; ‘Did the woman take the tablets during the past week?’ At each subsequent visit, almost all of the Selective iron women (98%) were measured for haemoglobin using the recommended HemoCue method and the same proportion of women in the Routine iron group were given iron tablets at each subsequent visit.
All analyses were performed on an intention-to-treat basis. Twin pregnancies (n=48 pairs) were included in the analysis because their numbers were similar in the two groups and their exclusion did not alter the results. For pregnancy outcomes, all women (n=4326) were included, whereas for birth outcomes, women with birth data (n=2258) were included. Differences in health indicators (fever, headache, cold/chills, nausea/vomiting, body aches, malaria) between the two iron groups at each subsequent visit (up to the fifth visit) during pregnancy were analysed by using binomial generalised estimating equations (GEE) with an exchangeable correlation structure. GEE takes into account the within person correlation in the setting of repeated measures.
Differences in continuously distributed birth outcomes (birth weight, duration of gestation, length of hospital stay) were analysed by using the two sample Student t tests. Categorical outcomes were analysed by using Pearson'sχ2 test or Fisher's exact test (in the case of cells with less than five cases). To estimate the risk ratios of the effect of iron, the binary birth outcomes (low birth weight (<2500 g), preterm birth (<37 weeks), caesarean section delivery, child and maternal ill-health or death at birth, negative fetal heart beat, delivery in a reference health centre, long hospital stay after birth (≥ 2 days) and unavailability of delivery data) were analysed by generalised linear models. The result estimates are presented with 95% CI. Statistical significance was set at p<0.05. STATA V.11 statistical software was used for the analyses.