Study setting and design
Women with singleton pregnancy and a recent diagnosis of GDM (< 29 gestational week, GW) were recruited for the study in the Helsinki Metropolitan area, Finland. The eMOM study had a combination of crossover and parallel study designs, starting after randomization (GW 24–28) with a two-week crossover phase (short-term effects). This was followed by a parallel design randomized controlled trial finishing to delivery (long-term effects). The current paper focuses on the crossover phase and the diets’ short-term effects on glucose and lipid levels.
The order of the intervention diets was randomized after eligibility and consenting to participation. The crossover period contained five study visits. The participants consumed their habitual diet during the run-in when entering the study and wash-out periods (between the intervention diets). The crossover schedule and participant flow through the study are presented in Fig. 1.
Participants and allocation
The women were recruited from the municipal outpatient maternal health clinics of Helsinki, Finland between December 2018 and March 2020, with the last participant completing the study in June 2020. Women were ineligible for participation if they had another type of diabetes than GDM, used medication that influenced glucose (e.g. oral corticosteroids or metformin/insulin) or cholesterol metabolism, had dietary restrictions that interfered with the intervention, they or the child’s father had other than European ethnicity, or had issues interfering with participation (e.g., inadequate Finnish language skills or substance abuse). The number of participants (aim n = 60) was decided based on the available resources and to provide a large enough number of participants at the end of the entire eMOM study, including the parallel study design that followed the crossover phase. Recruitment to the study took more time than planned and was stopped due to COVID-19 restrictions and reaching the end of the funding period.
To randomly assign participants to treatment arms, randomization with blocks of four was completed using a web page (URL: https://www.sealedenvelope.com), separately for nulliparous and multiparous women. The diet allocation was concealed in opaque sealed envelopes with continuous numbering. The envelopes were prepared by a researcher who did not meet the participants upon enrolment and allocation. The study personnel and participants were not blinded for diet allocation during the study because it was not possible considering the setting of the study. The data analyst was blinded for diet allocation during analysis.
If a participant needed pharmacological treatment (metformin or insulin) to maintain normal blood glucose, they were advised to contact their maternity health clinic for treatment and their participation in the study was discontinued.
Interventions
All foods and meals for the two 3-day crossover dietary periods were provided for the participants and their spouses/partners (to enhance adherence). Nutritionists planned the menus in collaboration with restaurant chef students and their teachers. The provided meals for both diets included 2200 kcal per day and participants got breakfast, two ready-made hot main meals, afternoon snack and evening snack daily. The goal macronutrient composition of both intervention diets is provided in Table 1. The full list of provided foods and meals and their total macronutrient contents are available in "Supplemental table 1.”
The Finnish Nutrition recommendations and mild carbohydrate restriction, which is currently (and was at the time of the study) recommended for GDM treatment in Finland (17), were the basis for the MCRD. The meals included mainly high fiber carbohydrates, for example brown rice, whole grain toast, oatmeal porridge and brown pasta. Meals contained around 0.5 kg of fruits and vegetables daily. Poultry and low-fat dairy products were the main protein sources in addition to two red-meat meals in a three-day period and protein from grains. Vegetable oils and vegetable oil spreads were the main fat sources. In addition, meals included fat from meat products and a few portions of nuts were provided.
The HND was characterized by Nordic ingredients, plant-based protein sources, and complex carbohydrates without restriction. The main sources of carbohydrates were Nordic whole grains (oats, barley and rye) as bread, porridge, pasta and cooked grains. Meals contained at least 1 kg of berries, fruits and vegetables daily with emphasis on Nordic species (e.g., bilberries, lingonberries, apples, onion, kale, and carrots). Legumes and processed plant-protein products (mainly fava beans) fermented low-fat dairy and fish (two fish meals for three days) were the main protein sources. Rapeseed oil and vegetable oil spread were the main sources of fat. In addition, small portions of nuts were included in some snack meals.
Table 1
Goal macronutrient characteristics of the MCRD and HND diets.
|
Moderately carbohydrate restricted
|
Healthy Nordic Diet
|
Energy, kcal
|
2000
|
2000
|
Protein, E%
|
20
|
20
|
Carbohydrates, E%
|
40
|
55
|
Sucrose, E%
|
< 10
|
< 10
|
Fat, E%
|
40
|
25
|
Saturated fatty acids, E%
|
< 10
|
< 10
|
Polyunsaturated fatty acids, E%
|
5–10
|
5–10
|
Dietary fiber, g
|
32
|
36
|
HND: Healthy Nordic Diet, MCRD: Moderately carbohydrate-restricted, E%: percentage of the total energy intake.
The participants marked their adherence to the diets in checklists. Checklist markings included eating time, if the provided meals were eaten fully, partially or not at all. Participants also marked if there were any changes to the meal plan, or if any extra items were consumed.
Outcomes
The primary outcome measure of this study was the percentage of time spent within the target blood glucose range (TIR, < 7.8 mmol/L) (17) measured by CGM (Freestyle Libre®, Abbott, USA) during two 3-day intervention diets. Secondary outcome measures were 3-day mean and standard deviation of glucose and change in blood lipid markers and homeostatic model assessment of insulin resistance (HOMA-IR) after the two 3-day intervention diets.
Data collection
CGM sensor
The CGM sensor (Freestyle Libre®, Abbott, USA) was attached to the participant’s upper arm at the first study visit. The diet intervention period started on the third day after CGM attachment to allow the sensor to adjust. The CGM sensor automatically measures the tissue glucose concentration every minute, and data are stored at a frequency of 15 minutes. The sensor is coupled with a hand-held reader, which shows the current tissue glucose concentration, a glucose curve of the last 8 hours and glucose measurement history. The sensor needs to be scanned with the reader at a minimum every 8 hours to collect the glucose data. The participants were instructed to scan the reader at least once during the daytime, before going to bed, and first thing in the morning.
Blood samples
Laboratory tests performed in conjunction with study visits included 10–12 hours fasting measurements of glucose metabolism (plasma glucose and insulin) and blood lipids (total cholesterol; high-density lipoprotein, HDL; and low-density lipoprotein, LDL; triglycerides, TG). Samples were drawn at the study center in the morning of the first day of each diet, and the morning following each diet end, in total four times. Venous blood samples were taken in a seated position with a light stasis and centrifuged at the survey site. Glucose samples were collected in 5 ml fluoride citrate vacuum tubes and analyzed by photometric hexokinase method. Insulin samples were collected in 5 ml serum gel tubes, centrifuged after collection, and analyzed with the immunochemiluminometric method. All blood samples for lipid markers (total cholesterol, HDL, LDL and TG) were collected in 5 ml li-heparin tubes, centrifuged, and analyzed with a photometric enzymatic method. All samples were analyzed by the Helsinki University Hospital laboratory (HUSLAB).
Questionnaires and baseline demographic data
All participants filled out a baseline questionnaire about demographic characteristics (age, educational level, social and work status, income level and profession), the current health status, habits (smoking, alcohol use; AUDIT (27)) and health history, use of medication and the history of diabetes and other illnesses in the close family. PA and activity level were self-rated using a standardized questionnaire commonly used in Finnish nationwide studies (28). Height and weight at baseline were measured by the study nurse and pre-pregnancy weight was collected from the maternity clinic patient card, in which standard pregnancy-related information is collected from and recorded for the pregnant women.
Data
Analysis and data preparation
The CGM glucose data that covered the days when the participant was following the diet protocols were extracted and preprocessed with Python 3 programming language (29). The first measurement day was defined as starting at 00:00 and the last measurement day ended at 23:59. TIR was calculated by counting the number of measurements that was ≤ 7.8 mmol/l and dividing it by the total number of measurements for each individual, creating a value representing the percentage of the total time. The individual participants 3-day mean, standard deviation of the 3-day glucose, the minimum and the maximum glucose values were calculated and added to a data frame (new variables) for further analysis of the various glucose measurements during the diets. Each individual’s 3-day CGM data length (i.e. number of observations) was inspected. If a participant had significantly lower amount of glucose observations than expected (> 20% data loss), their glucose curve was inspected. For two participants, there was significant loss of CGM data for longer time frames in one or both diet periods and these participants were excluded from analyses due to poor data quality.
All laboratory measures were entered to the database as double entries. All numerical data was checked for unusual values (i.e. outliers), and in case of unusual or missing values the original data was checked and corrected. A few cases of missing data were confirmed: missing blood sample data (one lipid panel and one fasting insulin), one participants parity and OGTT data was confirmed as not available and as such, not included in the respective analyses (highlighted in the results tables below). If participants had one or more family members (parents or grandparents) with any form of diabetes, this was coded with yes/no. Body mass index (BMI) was calculated by [weight in kilograms/(height in meters)2]. HOMA-IR value was calculated from fasting insulin and glucose using the following formula [fasting insulin (mU/L) × fasting glucose (mmol/L)/22.5]. Homeostatic model for beta cell function (HOMA β) was calculated using the following formula [20 × fasting insulin (mU/L)/fasting glucose (mmol/L) − 3.5] (30). Delta variables (Δ) to analyze postintervention changes were calculated for all blood test results and HOMA indices by subtracting the post-test value from the pre-test value and related to both periods of the crossover.
Statistical analyses
As this was a pilot study, no power calculation was made before the study. A per-protocol analysis approach was used, i.e., participants with incomplete CGM or checklist data were removed from the analysis (three participants, see CONSORT flow chart, Fig. 2). All data were analyzed for normal distributions by the Shapiro-Wilks test and by inspecting the distributions of the data with density and Q-Q plots. Paired t-test and paired Wilcox test were utilized to compare the intervention diets (each participant as their own control). A p value lower than 0.05 was considered statistically significant. All statistical analyses were performed with R programming language (version 4.0.5, The R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/) in RStudio (version 1.4.1106, “Tiger Daylily”, RStudio: Integrated Development for R. RStudio, PBC, Boston, MA, USA. URL: http://www.rstudio.com/). A sensitivity analysis was performed by removing one participant that was identified as having a larger difference in TIR (89% TIR in MCR, 94.8% TIR in HND, or 6.2% difference) between the diets compared to any other participant. Then, all analyses were repeated.
Adherence calculations and analysis
For full adherence to the diets, drinking water, coffee and tea was allowed, and minor switches between similar foods were approved e.g., pear instead of apple, cucumber instead of tomato. For adherence analysis, an adherence score was calculated. Each component of every meal in the diet checklist was given 1 point if consumed fully and 0 if omitted (apart from the exceptions mentioned above). If consumed partially, it was given 0.5 points. The points were summed separately for each day, and a percentage of adherence for each day was calculated based on the total points and the number of food items in all the meals. Spices or comparable items did not count towards the adherence points. If the participant consumed food items other than in the protocol or if meal items had been exchanged with non-protocol adhering items, these were given minus points (minus 1 to minus 0.25 points depending on the size of the serving) and redacted from the day’s total points. Given that the participants were of different body size, adding another serving of protocol-approved food items were not given minus points.