Risk factors for type 2 diabetes mellitus among out-patients in Ho, the Volta regional capital of Ghana: a case–control study

Objectives

The specific objectives of the study was to evaluate the risk factors for T2DM along the lines of socio-demographic, anthropometric and physical activity factors; dietary factors and biochemical factors among general population of patients who visit the OPD within the Ho municipality.

Study design

This study was an unmatched case–control study. In this study, patients diagnosed with T2DM receiving care at the two government hospitals of the Ho Municipality were recruited. Appropriate controls with similar ages who are also patients (but non-diabetic patients) receiving care (for conditions such as malaria, respiratory tract infections, gastroenteritis, anaemia, enteric fever, hypertension, intestinal worm infestations etc.) at these hospitals were recruited alongside with them. Both cases and controls were administered a questionnaire that comprises of standardized and validated tools. These tools include WHO STEPs instrument, GPPAQ and REAP. Additionally, the research participants were made to undergo physical examinations for weight, height and waist circumference. The participants were made to undergo laboratory testing of fasting venous blood to assess the biochemical factors of interest namely fasting blood glucose and fasting lipids. The distinction between those with the disease and those without the disease was based on the WHO criteria (for case definition for diabetes) [19]. Diabetes patients on lipid lowering drugs were excluded. Comparison of cases to controls for the various risk factors under different groupings namely; socio-economic status (SES), demographic and physical activity; dietary factors and biochemical risk factors was made. To be eligible for recruitment, the client must be living within the Ho Municipality; be within the ages of 35–65 years; must not be severely sick or bed ridden but be an otherwise normal ambulant out-patient. The diagnostic criteria for T2DM were as follows: Any patient aged at least 35 years but not older than 65 years with FBG of ≥7.0 mmol/L or a known diabetes mellitus patient diagnosed after the age of 35 years on oral anti-diabetic treatment. Any patient receiving care at the OPD of these two hospitals: Who is not a known diabetic and had laboratory confirmation of fasting blood glucose to be <7.0 mmol/L was recruited as control.

Study setting

Ho is the capital of the Ho Municipality located in the heart of the Volta Region of Ghana. It is also the regional capital. The Republic of Togo is on the Eastern border of the Ho Municipality and in the southern border is the Adaklu-Anyigbe District, in the western border is the South-Dayi district and the Hohoe Municipality is located on the northern border. The people of Ho are of the Ewe tribe however given its status as regional capital and municipal capital, it has attracted various groups of people from all over Ghana and beyond. It is therefore becoming a heterogeneous society.

There are two diabetic clinics in the Ho Municipality which provide care to the diabetic patients. These two clinics are also housed in the two major hospitals of the municipality that provide care to most patients in the municipality. The study population for this research was patients receiving clinical care at these two hospitals. Only otherwise, able and ambulant patients receiving care on out-patient basis at these two hospitals were recruited. In-patients on the hospital wards were not recruited as subjects in this study. The study used both incident and prevalent cases of T2DM. Both the cases and controls were patients receiving care at these two hospitals. This study was conducted in the Ho Municipality during the months of May to July, 2014.

The estimated prevalence of diabetes for Ghana is 3.35% [18] and the population size of Ho Municipal was 169,669 in 2010 according to unpublished data in the dhims of the Ghana Health Services. The highest reported prevalence of diabetes was in Kumasi among Pentecostal Christians which stood at 9.26% [17]. A similar case–control study among diabetes estimated the proportion of obesity among diabetics to be 58% whiles among the controls it was 33% [6].These values were used in the Epi-Info 7 software. Using these values with a power of 80%, confidence level of 95% and a ratio of 1:2 for cases and controls, a sample size of 46 cases with 91 controls (Total 137—Fleiss formula/method) is required. A total of 136 subjects were recruited for this study with 48 cases and 88 controls.

Data collection

Data on the cases and controls was collected using different tools merged together into one questionnaire. The main core of this instrument is the WHO STEPwise approach to surveillance (STEPS) which is a standardized tool design purposefully to be employed in the surveillance for risk factors for NCDs in member countries [20]. For the purposes of this study and easier analysis, some components of the STEPS will not be applicable and other segments will be too extensive so those parts were not included. Two additional standardized tools were incorporated. These were, a physical activity level assessment tool which is in the form of a questionnaire that is administered and it generates four categories of outcome. It is known as the general practice physical activity questionnaire (GPPAQ). It was developed by the London school of hygiene and tropical medicine [21]. The second is a more detailed dietary survey instrument. It is called the rapid eating and activity assessment for patients (REAP) developed for use in clinical settings [22].

Diagnosis of diabetes status was based on fasting plasma glucose levels as tested in the laboratory. Blood pressure measurement was done three times or as checked on three different occasions and the most consistent readings for systolic and diastolic blood pressures were taken as the true values. Information/data obtained from the patient interviews and the various data collection tools was cross checked with what is available on the patient’s hospital records, folder or insurance card whenever necessary and possible to ensure consistency.

Statistical analysis

Data obtained from this study was first examined for completeness and consistency as part of quality control check measures after which they were compiled into a spread-sheet format using the EpiData version 3.1 software which was subsequently exported into the Stata version 11 (StataCorp, College Station, TX, USA) for statistical analysis. Where missing data was encountered during the process of compiling the data, the client hospital records were retrieved to cross check. Where it was deemed problematic, those candidates were dropped out of the study. The tables summarized the relevant number that were finally analysed. However, for the GPPAQ, the information from this instrument was first summarized in the manner in which its protocol requires prior to its analysis in Stata. This summary was done on its web-based application which yielded four categories of outcome that were subsequently analysed in the Stata statistical software. Continuous variables like age were categorised.

Risk factors were analysed and reported by proportions among the following groups; sex; rural and urban dwellers; age groupings; and disease status. A lot of attention was given to analysis of the strength of association and interaction of key predictors like obesity, physical activity and triglyceride levels. Odds ratios (ORs) were obtained as the measure of association between the risk factors and T2DM in the bivariate analysis. Adjusted ORs were obtained following the multivariate logistic regression analysis whiles controlling for individually significant covariates in each group of exposures. There was a simple score adopted for the REAP questionnaire in which we simply score from zero to three for each of the 27 main question items on the REAP tool. The total for the responses was calculated as the score for individual respondents. These results were categorized into four groups and analysed.

Descriptive statistics

A total of 136 respondents participated in the study, consisting of 48 (35.29%) cases and 88 (64.71%) controls respectively. A total of 95 [39 (81.25%) cases and 56 (63.64%) controls] respondents underwent laboratory testing for fasting blood glucose and fasting blood lipid (total cholesterol, HDL cholesterol and triglycerides). By gender, there were more women (58.09%) than men (41.91%). The youngest subject was 35 years old and the oldest was 62 years old. The mean age of the respondents was 47.68 (sd = 6.98) years. Majority of the respondents were in the 40–49 years age group 63 (46.32%) of the total.

Socio-demographic, anthropometric and physical activity factors

The place of residence of the respondents was grouped into urban and rural depending on the level of development of the area and whether it is a sub-district of the Ho Municipality. Majority (57.35%) of participants were grouped into places considered as urban as against 42.65% of rural dwellers. Interestingly, urban dwellers were in the majority (75%) among the cases, compared to 25% of rural dwellers. The bivariate analysis comparing the effect of the place of residence on risk of developing T2DM gave an OR of 3.29 (p < 0.0018) comparing urban dwellers to rural dwellers. This relationship continues to hold true when SES or body mass index (BMI) are controlled for individually (p < 0.05). Nevertheless, when physical activity level (GPPAQ) scores are controlled, the OR changes to 0.90 (p < 0.852).

Information obtained from the subjects on educational level, income level and type of work or employment status was used to derive a SES score for each participant. The scores were further grouped into three to represent low, middle and high strata of SES. Respondents in the middle and high strata of SES were compared to those in the low SES as a reference group, and the following results were obtained. Those in the middle SES had an OR of 2.97 (p < 0.007) while those in the high SES had an OR of 0.9 (p < 0. 929). After controlling for the effects of place of residence, GPPAQ score and BMI the OR for the middle SES increase to 5.03 (p < 0.003).

The number of days within the week in which subjects consumed fruits was obtained and categorized into two. An OR of 3.43 (p < 0.004) was observed comparing those who consume fruits 5 or more days in a week to those who consume fruits up to 4 days in a week. Adjustment for other dietary components (servings of fruits per eating and REAP dietary score) yielded an OR of 4.08 (p < 0.003). Besides the frequency of fruit consumption, the servings of fruits consumption per eating was also obtained and categorized into three. Those who eat four or more servings of fruits; those who eat two to three servings of fruits; and those who eat one or less servings of fruits. Those who eat four or more servings were the reference group. The ORs were 3.22 (p < 0.012) and 3.75 (p < 0.017) 0.017 respectively. After adjusting for the effect of the REAP dietary score and frequency of fruit consumption per week on the development of T2DM, the ORs became 3.15 (p < 0.02) and 5.76 (p < 0.005) respectively. Subjects with REAP scores of 31–40; 41–50 and ≥51 compared with those having scores ≤30 had ORs of 6.09 (p < 0.035); 5.86 (p < 0.046) and 7.34 (p < 0.055) respectively after adjusting for fruit consumption.

The levels of total cholesterol were categorized as shown in Table 2 with the ≤5.2 mmol/L category being the reference group. Respondents with total cholesterol values 5.2–6.2 mmol/L and those ≥6.3 mmol/L had ORs of 2.27 (p < 0.131) and 5.00 (p < 0.012) respectively. After adjusting for the effect of BMI and HDL cholesterol, those in the 5.2–6.2 mmol/L group had an OR of 4.09 (p < 0.033) while those with ≥6.3 mmol/L cholesterol values had an OR 10.67 (p < 0.002). For triglycerides, the cut-off point was 1.7 mmol/L and those with fasting triglyceride levels of ≥1.7 mmol/L had an OR of 6.71 (p < 0.088). HDL cholesterol was grouped into three. The best group (<1.0 mmol/L for men and <1.3 mmol/L for women), better group (1–1.3 mmol/L for men and 1.3–1.5 mmol/L for women) and poor group (> 1.3 mmol/L for men and > 1.5 mmol/L for women). The reference group was the best group and the following ORs were obtained; 0.81 (p < 0.778) and 1.99 (p < 0.121) for the better and poor groups respectively. Interestingly the situation changes dramatically for the poor group of HDL cholesterol in the presence of total cholesterol. The OR then becomes 4.74 (p < 0.007).

Limitations of the study

Analysis on differences in the risk factors according to ethnic groupings could not be performed because majority of our participants were of one ethnic group (Ewes). Genetic predisposition by way of family history of diabetes mellitus was not assessed. A total of 136 candidates were recruited instead of the 137 candidates as required by the sample size calculation using the Fleiss formula/method in the EpiData 7 Software.

For majority of the cases, they were old T2DM patients so recall bias is a possibility. However for those cases who were newly diagnosed or diagnosed by the work of the research team, this is less likely. However there was no separate sub-group analysis between incident and prevalent cases to draw a conclusion on these aspects. This is because the incident cases were too few.