Effect of age and sex on the associations between potential modifiable risk factors and both type 2 diabetes and impaired fasting glycaemia among West African adults

The WHO’s Stepwise Approach to Surveillance (STEPS) survey method was used to collect individual population-level data between 2006 and 2013 from five different West African countries, namely, Benin, Burkina Faso, Ghana, Liberia, and Mali. The STEPS survey is a standardised instrument used to collect information on NCD risk factors in WHO member states [17]. All WHO member states eligible to participate in the survey decide what data to collect based on their needs and interests, as well as available resources and the capacity to implement the survey [17]. The STEPS survey comprises three components, which include data from eight behavioural, biological, and biochemical factors contributing to the burden of NCD, including T2DM. The information is collected through a questionnaire (Step 1); physical examination (Step 2); and biochemical measurements (Step 3). Each country obtained ethical approval from its respective ethics committee. Informed consent was obtained from each participant. Multi-stage cluster sampling was used to randomly select participants [18]. Out of 14 West African countries invited to participate, five countries responded and provided their WHO STEPS data.

The total sample included 16,845 participants before data cleaning. After excluding observations that had incomplete, inconsistent and invalid information, records on 15,520 participants were retained for analysis. To ensure data comparability between countries, only participants aged between 25—64 years were included. The survey response varied across countries from 95.2% (Mali) to 99.4% (Ghana). Detailed information about the study design is available from the reports of participating countries on the WHO STEPS websites [19]. Missing data were generally scant in all countries. We conducted a sensitivity analysis to ascertain the potential effect of data not being missing at random. However, owing to the low proportion of missing values, more complex approaches to missingness such as multiple imputations were not warranted. Details of these sensitivity analyses have previously been reported (see Issaka et al. [20]). Before our sensitivity analysis, we checked for internal consistency and reliability of the data by ensuring that no records with incompatible variable values were included and that all records of variables values agreed with each other. To ensure all values were realistic, we excluded implausible values using the WHO-recommended cut-off values [18]. A measure of HBP was not available for data from Ghana, while hip circumference was not collected in Burkina Faso meaning that the WHR could not be calculated for this country.

All definitions followed the WHO-recommended cut-off values [21]. IFG was defined as 6.1 – < 7.0 mmol/L (110 – 125 mg/dL). T2DM was defined as a fasting plasma glucose reading of ≥ 7.0 mmol/L (> 126 mg/dL). Hypertension was defined as diastolic blood pressure ≥ 90 mmHg and/or systolic blood pressure ≥ 140 mmHg [21]. Participants who reported taking blood pressure lowering medication were classified as having hypertension regardless of their blood pressure measurements in the survey. Normal BMI was defined as 20 – 24.99 kg/m², overweight was defined as BMI of 25 – 29.99 kg/m², and obesity as BMI ≥ 30 kg/m². Elevated WC was defined as ≥ 80 cm for men and ≥ 94 cm for women. Elevated WHR was defined as ≥ 0.90 for men and ≥  0.85 for women. Elevated WHtR was defined as WHtR > 0.5. Fruit and vegetable consumption was defined as inadequate and adequate. Daily smokers were defined as those who currently used tobacco daily, while current alcohol drinkers were defined as those who have drunk alcohol at least once over the last 30 days. Those who drank alcohol every day per week were considered heavy drinkers. Physical activity (low, moderate, or high) was categorised according to self-reported answers to questions from the Global Physical Activity Questionnaire (GPAQ). Owing to the low level of fruit and vegetable consumption in this sub-population, adequate fruit and vegetable consumption was defined as two servings per day instead of the WHO-recommended five servings per day [22]. Among the sociodemographic variables, employment status was dichotomised as either employed or unemployed while educational status was categorised as: none; primary, and secondary/tertiary. Sex was coded as Male = 0 and Female = 1.

Stata 17.0 was utilised for all analyses. All analyses were adjusted for the clustered sampling design used, with data weighted to the age and sex profile of the African standard population [23]. Across all analyses, a p-value of p < 0.05 was considered statistically significant. Variables were described using simple percentages, means (reported as mean ± standard deviation), and frequencies as appropriate. Student t-tests tests were used to compare different group means of male and female participants. Chi-square was used to compare the categorical variables. Multinomial logistic regression was used in a pooled analysis, separately for males and females, to ascertain the relative risks (RR) and 95% confidence intervals (CI) of each risk factor modelled as categorical variables based on established cut-points and the same outcomes. Associations between the traditional risk factors and both T2DM and IFG were assessed in crude analyses, and after adjusting for confounding factors. The confounding factors include age, sex, education, and profession, and were considered as all five countries had data for them. We tested whether the association between these traditional risk factors and both T2DM and IFG varied between males and females and across different age groups by adding interaction terms for sex and age as continuous variables. The testparm command in Stata was then used to ascertain the presence of statistically significant interaction simultaneously across the categories of the traditional risk factors and both T2DM and IFG as outcomes.

The current study has several strengths, including the large sample size from five different West African countries. This ensured greater statistical power to detect age and sex interaction effects between potentially modifiable traditional risk factors and both T2DM and IFG. The study, however, has several limitations, which need to be considered when interpreting the results. First, because the study is cross-sectional in design, results do not imply causal relationships between these traditional risk factors and T2DM. Secondly, only the traditional risk factors that were assessed in all five countries were analysed. Therefore, important risk factors, such as high-density lipoprotein cholesterol levels, could not be included in this analysis, though studies have shown this factor to interact with sex [8]. Thirdly, since an OGTT was not used to define diabetes and pre-diabetes, the results may differ from those where this method was used, especially given the sex-specific impact of a glucose load due to differences in body size between males and females [31]. Fourthly, fruit and vegetable consumption measures were not coded as per the WHO guidelines of five servings per day due to the low level of fruit and vegetable consumption in this sub-population and may constitute a limitation when compared to other studies. Lastly, the different years that the survey data were collected may have introduced some bias, however, this is not an important limitation for this study given the focus on associations between risk factors and health outcomes. These limitations notwithstanding, the findings from the current study have important policy implications.

Since the associations between the traditional risk factors and both T2DM and IFG appear to vary minimally based on age or sex, policies and interventions do not need to be tailored to different West African populations based on age or sex. This is particularly advantageous given the low-income context in West Africa, with population-wide interventions likely to be both more cost-effective and simpler to implement. While, in general, smoking and alcohol are more prevalent among men, obesity and physical inactivity are more prevalent among women in West Africa. Any policies targeting these risk factors should consider socio-cultural factors and beliefs [51]. These may include i) the commonly held belief in much of Africa that being overweight is an outward manifestation of high socioeconomic standing, prosperity, and beauty, as well as good health among females and the preference for central obesity among some affluent men [52, 53]; ii) the fact that physical activity among women is often discouraged in most countries as it is culturally considered to be undesirable and unattractive and associated with a masculine physique [51], and iii) the fact that physical activity is usually not viewed through a health lens for men, but through the lens of sports [54]. Policies should target the persistently low levels of awareness regarding the importance of fruit and vegetable consumption, as well as the globalisation of food markets, particularly concerning alcohol and tobacco industries. Globalisation has been shown to exacerbate the use and increased the ease of access to alcohol and tobacco use among young adults in Africa [55].