Detection and initial management of gestational diabetes through primary health care services in Morocco: An effectiveness-implementation trial

Background Gestational Diabetes Mellitus (GDM) testing and management in Morocco is associated with delays resulting in late commencement of treatment. To reduce delays and to increase access of women to GDM care, a country-adapted intervention targeting primary health care providers was designed to test the hypothesis that detection and initial management of GDM at the primary level of care improves newborn outcomes in terms of lower birthweights and less cases of macrosomia and impacts on maternal weight gain, glucose balance and pregnancy outcomes. Materials and methods We conducted a cluster randomized controlled trial in two districts of Morocco. In each district, 10 health centers were randomly selected to serve either as intervention or control sites. Pregnant women attending antenatal care in the study facilities were eligible to participate. At the intervention sites, women were offered GDM screening by capillary glucose testing following International Association of Diabetes in Pregnancy Study Groups/WHO criteria. Women diagnosed with GDM received counselling on nutrition and exercise and were followed up through their health center whereas at control facilities routine practice was applied. Primary outcome was birthweight and secondary outcomes maternal weight gain, glucose control and pregnancy complications. We further assessed GDM prevalence in the intervention arm. Statistical analysis was performed on 210 recruited women. Continuous variables were reported using means while categorical variables using frequencies with tests of independence applying chi-squared tests. Differences of outcome variables between the two groups were estimated by mixed-effects regression models and effect sizes adjusted for confounders. The trial is registered under NCT02979756 at ClinicalTrials. gov. Results GDM prevalence reached 23.7% in Marrakech. Birthweight in the intervention group was 147grams lower than in the control group (p = 0.08) as was the proportion of macrosomes (3.5% versus 18.4%; p< 0.001). In the intervention arm, women did two times more followups than at control sites (p = 0.001) and mean follow-up intervals were shorter (11.3 days versus 18.7 days; p < 0.001). Overall, 30% more fasting blood sugar values were balanced (p = 0.005) and mean weekly maternal weight gain 49 grams lower (p = 0.032) in the intervention group. More women from control facilities had a delivery complication whereas more newborn complications were observed in women from intervention facilities. No difference between the two groups existed regarding mode of delivery and mean gestational age at delivery. One of the main limitations of the study was the Hawthorn-effect at control sites that might have led to an underestimation of the effect size. Conclusion A high GDM prevalence in Morocco calls for a context-adapted screening and management approach to enable early interventions. GDM detection and care through antenatal care at primary health facilities may have positively impacted on newborn birthweight but findings are inconclusive. Results of this study will contribute to the decision on a potential upscaling of the intervention in Morocco. Future research could examine long term metabolic changes including diabetes type 2 in the cohort of women and their children. As part of an effectiveness implementation study, we conducted a cluster randomized controlled trial in the region of Marrakech-Safi located in central Morocco and covering part of the Atlas Mountains. Two districts, Marrakech and Al Haouz were chosen for this implementation trial with the former being mainly urban and the latter being a predominantly rural district with limited geographical access to health care services. Around 1.95 million people live in the area, where 92 health centers and 53 dispensaries provide primary health care to the population.[12] We chose as unit of randomization the health center (cluster) to limit contamination between the different arms. To select the health facilities for the study, we draw from the list of health centers provided by the district directorates all those that performed on average at least 30 monthly ANC consultations (new cases). These were then continuously numbered in each district and a random number generator used to select 10 health centers per district, the first five serving as intervention and the remaining five as control facilities. In the selected facilities, all pregnant women of consenting age attending ANC and being newly diagnosed with GDM were eligible to participate in the trial. Pregnant women with a known diabetes type 1 or 2 were excluded. ANC providers of the intervention facilities (two providers per site) received a 1-day training on GDM screening and management. The training targeted the application of new GDM algorithms for screening and management that were based on the latest FIGO consensus recommendations [10], the preparation and conduct of a 75g oral glucose tolerance test (OGTT) and the provision of advice on nutrition and physical exercise. All ANC services at the intervention sites received the necessary equipment for GDM screening including glucometers, test-strips, 75g OGTT solutions as well as brochures and guidelines on nutrition for women affected by GDM that had been developed by the Moroccan Ministry of Health. From 14th November 2016 (World Diabetes Day) onwards, women attending ANC at the intervention sites were offered GDM screening by capillary glucose testing. Applying the IADPSG and WHO criteria as adopted by FIGO in their latest consensus recommendations for the diagnosis of GDM [8,10,13], women tested positive for GDM at the intervention facilities either by a fasting blood glucose (FBG) before 24 weeks or by a 75g OGTT in the second trimester received nutritional counselling by the ANC nurse and were followed up weekly to twice monthly up to eight weeks post-partum through the health center [14]. Cut-off values for the diagnosis of GDM were 0.92–1.25g/l (5.1−6.9 mmol/l) for fasting glucose, or 1.8g/l (10 mmol/l) and above one hour, or between 1.53 and 1.99 g/l (8.5–11.0 mmol/l) two hours after ingestion of 75g glucose [10]. If during follow-up glucose values exceeded internationally recommended targets for fasting or postprandial values [10,15], women were sent for consultation to the diabetes referral center. At the control facilities, screening and initial management of GDM followed routine practice using national recommendations.[16] According to latest national guidelines, pregnant women detected with a fasting glucose value of 0.92g/l in the first trimester or at their first ANC visit should be diagnosed with a GDM without a repeat diagnostic test being required. If the obligatory first trimester glucose test was not done or normal, a 75g OGTT is advised between 24 and 28 week gestational age (GA) following the international IADPSG thresholds for diagnosis. Providers were asked to recruit every woman with GDM until they reached at least 8 GDM patients who consented to participate in the study. The last woman in the intervention arm was included on 6th June 2017 and in the control arm on 14th November 2017. The trial is registered under {"type":"clinical-trial","attrs":{"text":"NCT02979756","term_id":"NCT02979756"}}NCT02979756 at ClinicalTrials.gov and was first posted on 2 December 2016, two weeks after starting the enrollment of study participants due to administrative reasons. The authors confirm that all ongoing and related trials for this intervention are registered. Primary outcome measure was the difference in mean birthweight between the two arms, while secondary outcomes were maternal weight gain and glucose levels during follow-up, pregnancy related outcomes including mode of delivery, presence or absence of obstetric (prolonged labor, pre/eclampsia, shoulder dystocia and other) or newborn (respiratory distress, hypoglycemia and other) complications and prevalence of GDM at the health centers using as numerator the number of women tested positive for GDM and as denominator the number of pregnant women tested. Maternal weight (pre-pregnancy weight as reported by the mother and measured during ANC/ follow-up visits) as well as information on glucose levels (measured at the health center or in the private sector) were documented by the treating nurse on a preformatted data collection form. Information on maternal and newborn outcomes (birthweight, complications during delivery, Apgar) relied on information provided by the ANC nurses and was verified for 91% (n = 191) of women in the respective hospital records. For home deliveries (5.2%), birthweight had to be estimated based on the weight of the newborn taken post-partum at the health center (between 1 and 26 days after delivery). Birthweight was estimated by two independent researchers using the 50th percentile of the weight change nomograms developed by Paul et al. [17]. We investigated the hypothesis that the primary outcome, mean birthweight, differs in both groups. Using previous studies on birthweight differences of newborns of GDM affected mothers receiving different treatment schemes [18,19], we assumed that mean birthweight of newborns in the intervention group would be 300 grams lower than mean birthweight in the control group (3400 versus 3700 grams with a standard deviation of 500 grams), due to earlier detection and treatment of GDM. For our sample size calculation, we applied an intra-class correlation coefficient of 0.1 and assumed 20 equally sized clusters. Under 80% power and 5% alpha, this yielded a sample size of 75 GDM affected women per treatment arm, or 7.5 per health center, which we rounded up to 8. Based on a previously reported GDM prevalence of 8,2% in Morocco [20], we assumed that we would need an average cluster size of at least 90 women tested for GDM per health center to detect 8 women with a GDM. Individual patient data including age, gestity, parity, presence of previous obstetric risk factors, last menstrual period, weight and height, GDM test results and date of diagnosis, treatment, information on referral and follow-up (glucose test results, weight, treatment) as well as routine monthly facility data were collected by the ANC nurse at the health center and reported on pre-formatted data collection forms. The research coordinators of the two districts visited the health centers in 4–6 weekly intervals to collect the information and clarify any inconsistencies. Data on outcomes was collected at the health center and at the place of delivery. All data was anonymized and double entered into a preformatted excel file, cleaned and later converted into STATA version 14 and R version 3.4.2 for further analysis. Statistical analysis was performed on all 210 recruited women irrespective of gestational age at diagnosis and their follow-up adherence based on the intention-to-treat principle. Preterm deliveries (gestational age below 37 weeks; n = 8) were excluded from birthweight analyses. Continuous variables were reported using means and standard deviations or standard errors while categorical variables described by frequencies and percentages. Differences of outcome variables between the two groups were estimated using the mixed-effects regression model by applying a linear mixed model for continuous outcomes, logistic for categorical outcomes and Poisson to estimate number of balanced blood sugar values, using the total number of measurements per patient as an offset variable. Most models employed two levels of random effects, patients and health centers. The analyses of maternal weight gain and time interval between follow-up visits were based on disaggregated within-person measurements and employed three levels: within patients, between patients and between health centers. Effect sizes were adjusted for maternal age, gestity and residence. All models were estimated using Restricted Maximum Likelihood (REML); the logistic and Poisson models additionally employed the Laplace approximation. The use of random intercepts resulted in a compound symmetry correlation structure. Tests of independence based on two-by-two contingency tables were performed as chi-squared tests with Rao and Scott’s (1984) correction to account for clustering within health centers [21]. All women included in this study were asked for written consent to participate. They were free to decline or stop their participation at any time without any consequences for their care. The study was approved by the institutional review board of the Institute of Tropical Medicine (Reference 1086/16) and the Ethics Committee of the University Hospital (Registration B300201628508) in Antwerp, Belgium and by the Ethics Committee for Biomedical Research at Mohammed V University, Rabat (Dossier 83/16).