Shorter Height is Associated with Diabetes in Women but not in Men: Nationally Representative Evidence from Namibia

Objective: This study aimed to test the hypothesis that attained adult height, as an indicator of childhood nutrition, is associated with diabetes in adulthood in Namibia, a country where stunting is highly prevalent. Methods: Data from 1,898 women and 1,343 men aged 35 to 64 years included in the Namibia Demographic and Health Survey in 2013 were analyzed. Multiple logistic regression models were used to calculate odds ratios (ORs) and 95% CIs of having diabetes in relation to height. The following three models were considered: Model 1 included only height, Model 2 included height as well as demographic and socioeconomic variables, and Model 3 included body mass index in addition to the covariates from Model 2. Results: Overall crude diabetes prevalence was 6.1% (95% CI: 5.0-7.2). Being taller was inversely related with diabetes in women but not in men. In Model 3, a 1-cm increase in women’s height was associated with 4% lower odds of having diabetes (OR, 0.96; 95% CI: 0.94-0.99; P = 0.023). Conclusions: Height is associated with a large reduction in diabetes in women but not in men in Namibia. Interventions that allow women to reach their full growth potential may help prevent the growing diabetes burden in the region. Data was extracted from the 2013 Namibia Demographic and Health Survey (DHS), a nationally representative population‐based household survey. DHS surveys are designed to collect data on fertility, family planning, and maternal and child health to assist countries and researchers in assessing changes in population, health, and nutrition 27. The 2013 Namibia DHS was implemented by the Ministry of Health and Social Services together with the Namibia Statistics Agency and the National Institute of Pathology. The sampling frame used for this survey was the preliminary frame of the Namibia Population and Housing Census conducted in 2011. For the sampling approach, each of the 13 administrative regions was divided in small enumeration areas covering the entire country. In a two‐stage process, 20 households were randomly selected in every urban and rural cluster. In addition to demographic, socioeconomic, and health data, the 2013 Namibia DHS measured the prevalence of high fasting plasma glucose (FPG) and collected anthropometric measurements of weight and height. A total of 9,849 households were successfully interviewed, yielding a household response rate of 97%. Anthropometric characteristics and FPG were investigated in a subsample of women and men aged 35 to 64 years. Of the 2,584 women and 2,163 men eligible for the FPG test, 74% of women and 63% of men had their plasma glucose measured 27. In our analysis, we included only participants with full information on our exposure, outcome, and covariates, yielding a total study population of 1,898 women (73% of eligible women) and 1,343 men (62% of eligible men) aged 35 to 64 years. Our outcome was diabetes using WHO cutoff points. FPG values ≥ 7.0 mmol/L (126 mg/dL) were defined as diabetes 33. Plasma glucose was measured in capillary whole blood after an overnight fast. Although the WHO recommends venous plasma glucose as the standard method for measuring glucose concentrations in blood, capillary sampling is widely used, particularly in low‐resource settings. Fasting values for venous and capillary plasma glucose are considered to be identical 33. All respondents included in our study confirmed that they had fasted for at least 8 hours prior to measurement. In addition to a blood test, participants were asked whether they were taking medication for diabetes (“Are you currently receiving prescribed medication such as insulin for your high blood glucose or diabetes?”). Individuals reporting use of drugs for diabetes were classified as having diabetes, irrespective of their biomarker values. Our outcome for diabetes was treated as a dichotomous variable (diabetes/no diabetes). Our key exposure was attained adult height as a continuous variable in centimeters. Adult height was measured once in a standing position using a Shorr height board (Weight and Measure, LLC, Maryland, USA). Measurements were performed by trained staff following a detailed field manual. We included sex, age (continuous), residency (urban, rural), educational attainment (level attained at the time of the survey), and household wealth quintile as covariates to control for potential confounding in our analysis, in addition to body mass index (BMI) as a potential mediator. The household wealth index is a composite measure of a household’s cumulative living standard. It is calculated using data on a household’s ownership of selected assets and is typically based on 25 to 50 survey questions, such as materials used for housing construction and types of water access and sanitation facilities. The wealth index is generated based on principal components analysis (as per standard DHS methodology) and places individual households on a continuous scale of relative wealth. In the DHS, interviewed households were separated into five wealth quintiles (five being the wealthiest) 27. BMI was grouped into four categories, thin (BMI < 18.5 kg/m2), normal (BMI 18.5‐24.9 kg/m2), overweight (BMI 25‐29.9 kg/m2), and obesity (BMI ≥ 30 kg/m2), in accordance with the WHO classification 34. Data on waist circumference or ethnicity were not available in the DHS. The main objective of our analysis was to investigate the relationship between attained adult height and diabetes, controlling for potential confounders. Three different multiple logistic regression models were fitted. The models considered were Model 1, including only height to examine the unadjusted effect of height; Model 2, including height along with demographic and socioeconomic variables (age, sex, residency, education, wealth); and Model 3, including BMI along with the covariates of Model 2. We tested our assumption of a linear relationship between attained adult height and diabetes by adding higher‐order polynomial terms of height into the models, but none of the associations became significant at the 0.05 level in the pooled data of women and men with the exception of Model 1. We therefore kept height in the models as a linear term. All regression analyses were conducted in the pooled study sample, as well as stratified by sex to determine sex‐specific associations. Odds ratios (ORs) with 95% CIs for correlates of diabetes were estimated. Sample weights were used as provided by the DHS for all descriptive statistics. We clustered standard errors at the enumeration area level to take into account spatial correlation between respondents. We conducted a wide range of supplementary analyses to generate additional confidence in the robustness of our findings. First, we used an alternative specification of our exposure. We transformed height into a categorical variable based on quartiles, in which Q1 represents the shortest and Q4 the tallest height category. Second, we included BMI as a continuous variable in our model to fully adjust for BMI instead of the BMI categories thin, normal, overweight, and obesity. Third, we used alternative specifications for age and BMI to model the nonlinear relationship of age and BMI with diabetes risk. We included quadratic and cubic terms in age as well as BMI in our models. Fourth, we restricted our sample to workers and included occupation as a covariate in our models to further account for lifestyle differences. We used the occupation categories as provided by the DHS, which are based on the International Standard Classification of Occupations. Fifth, we excluded pregnant women (n = 35) from our analytical sample because pregnancy may affect blood glucose. Sixth, we modeled our outcome with a log link function in Poisson regression models. Seventh, we graphically assessed the relationship between the proportion of children under 5 years of age who were stunted and mean attained adult height by region in Namibia. Lastly, we conducted a placebo test. We restricted our analytical sample to married couples and ran four additional models in which we regressed the following: (1) wife’s diabetes status on wife’s height, (2) wife’s diabetes status on husband’s height, (3) husband’s diabetes status on wife’s height, and (4) husband’s diabetes status on husband’s height. We hypothesized that if adult height could be interpreted as a true (causal) exposure of diabetes, the coefficients from Models 2 and 3 would likely be substantially smaller relative to the coefficients obtained from Models 1 and 4, respectively. Conversely, if the coefficients from, for instance, Models 1 and 2 would be similar in effect size, a possible causal interpretation of Model 1 would appear problematic and suggest possible residual socioeconomic confounding or assortative mating 35. Stata (version 15.0; StataCorp, College Station, Texas) was used for all statistical analyses. This study was considered exempt from full review by the Harvard T.H. Chan School of Public Health Institutional Review Board, as the analysis was based on an anonymous public use data set with no identifiable information on the survey participants.