ABE JEROME BLOG

Study design

This was a cross-sectional survey to determine prevalence of diabetes and IFG among people attending six health centres in six different districts of Luanda (Angola), followed by a case-control study to assess the factors associated with IFG and diabetes in a subgroup of subjects not receiving treatment for diabetes.

Setting

The study was conducted in the urban area of Luanda (Angola), with a population of approximately 7 million people. According to national authorities, the capital city of Angola has the highest diabetes mortality rate in the country (3.4%)9. One health center was randomly chosen in each of the six districts. People attending the six health centres between August 2018 and November 2018 were evaluated for inclusion in the study.

Participants

The study population included adult men and women older than 18 years attending the six health centres between August 2018 and November 2018. Pregnant women and people not fasting for at least 8 hours were excluded from the study.

Cross-sectional survey

All participants were included in the cross-sectional survey to determine the prevalence of IFG and diabetes. Fasting blood glucose (FBG) levels were measured by professional nurses using a glucometer (Infopia, South Korea). Diabetes measurements and definitions were based on WHO guidelines10. IFG was defined as FBG levels between 110–125 mg/dl. Diabetes was defined as FBG levels ≥126 mg/dl or a self-report of previous diagnosis of diabetes by a health care professional or currently receiving treatment for diabetes.

Case-control study

A subgroup of participants was selected for the investigation of factors associated with IFG and diabetes among those who were not receiving treatment for diabetes, as follows. All subjects with IFG were included. The chosen ratio of participants with IFG, diabetes and without IFG /diabetes was 1:1:1. For each participant with IFG, the next participant without IFG /diabetes was also included in the second step. Participants with diabetes were included in the second step and "linked" with the closest (past or next) participant with IFG, to warrant the achievement of the set 1:1:1 ratio. This approach was chosen to overcome the logistics issues of evaluating the factors in all participants of the cross-sectional survey.

A set of information of interest for IFG/diabetes (demographics, clinical parameters, clinical information, diet habits and physical activity) was collected by professional nurses using a case-report form. Demographics included age, sex, marital status, employment, education, number of children and number of people living at home. Clinical parameters included weight, body mass index (BMI), waist circumference, systolic blood pressure (SBP), diastolic blood pressure (DBP) and heart rate. Clinical information included having already heard about diabetes, having glycaemia measured before, losing weight, and feeling polyuria, polydipsia and weakness. Diet habits included eating vegetables, fruits, free-sugars food, adding refined salt in meals and drinking alcohol or free-sugars beverages daily. Tobacco use was self-reported. Information on physical activity included whether the subject usually performed physical activity for at least 30 minutes per day, and how long the subject rem ained seated daily (<2 hours, 2–5 hours,>5 hours). Demographics, clinical information, diet habits and physical activity were self-reported, while clinical parameters were measured by a professional nurse. All data were collected by a professional nurse before participants went into the doctor's consultation room. Data collection was coordinated and supervised by one research assistant.

Sample size

The sample size was calculated according to the primary aim of reporting the prevalence of diabetes and IFG in the cross sectional survey. Assuming a proportion of participants with IFG between 60 and 90 per 1,000 subjects, some 963 to 1,400 participants needed to be enrolled to estimate a 95% confidence interval (CI) not wider than 30 per 1,000 subjects. Assuming a proportion of participants with diabetes between 80 and 120 per 1,000 subjects, some 1,257 to 1,803 participants needed to be enrolled to estimate a 95% CI not wider than 30 per 1,000 subjects. The largest of the calculated sample sizes (1,803 participants) was chosen.

Statistical analysis

In the cross sectional survey, the proportions (with 95% confidence interval, CI) of participants with IFG, diabetes or without IFG/diabetes were estimated in all sample.

In the case-control study, factors associated with IFG and diabetes were investigated in the subgroup of participants, who were divided in three study groups (no IFG/diabetes; IFG; diabetes). Continuous data were reported as median and interquartile range (IQR), while categorical data as number and percentage. Continuous data were compared among study groups using Kruskal-Wallis test, and categorical data using Chi-square test or Fisher's exact test, as appropriate.

Multivariable regression was performed to identify independent predictors of study group (no IFG/diabetes; IFG; diabetes) among a set of clinically relevant variables. These included age, sex, marital status, employment, weight, BMI, waist circumference, SBP, DBP, having glycaemia measured before, polyuria, weight loss, polydipsia, weakness, eating vegetables, eating fruits, consuming free-sugars food and beverages, alcohol assumption, adding refined salt in meals, doing physical activities for at least 30 minutes and time spent seated. Some of them (DBP, BMI, waist circumference, marital status, employment) were not included in the initial model due to collinearity with SBP, weight or age. Eating free-sugars food and drinking free-sugars beverages conveyed the same meaning, thus were combined in a single variable (named "consuming free-sugars food/beverages"). As proxy for sedentary habits, time spent seated was preferred to "doing physical activities for at least 30 mi nutes". Model selection was performed by AIC reduction. Owing to the ranked outcome, the ordered logistic regression model was initially chosen. The proportional odds assumption was checked graphically because statistical tests for this purpose have been criticized to be un-conservative11. The proportional odds assumption was not satisfied for some variables, thus a generalized ordered logit (partial proportional odds) model was estimated12. Effects sizes were reported as odds ratios (OR) with 95% confidence intervals (CI), where OR1 indicated the odds ratio for IFG/diabetes vs. no IFG/diabetes, and OR2 indicated the odds ratio for diabetes vs. no diabetes. The analysis estimated two different odds ratios (OR1 and OR2) for the explanatory variables violating the proportional odds assumption, while equal odd ratios (OR1 = OR2) were reported for all other explanatory variables not violating the assumption. Missing data were not imputed. All test were 2-sided and a p-value less t han 0.05 was considered statistically significant. Statistical analysis was performed using R 3.5 (R Foundation for Statistical Computing, Vienna, Austria)13.

Ethics

This study was approved by the National Public Health Directorate of the Ministry of Health of Angola and by the Ethics Committee of the Ministry of Health of Angola (number 21/2018). Each participant signed a full informed consent form. All methods were performed in accordance with the relevant guidelines and regulations. The study used anonymized data and no identifiable data were collected.