We found that in men and women the risk of future type 2 diabetes was lower by more than 30% for each 10 cm difference in height when accounting for common diabetes risk factors. For components of height, specifically leg length was inversely associated with risk in men and women. These findings are partly in line with the few previous prospective studies on height components and diabetes risk [8, 9]. In the Shanghai Women’s Health Study and the Shanghai Men’s Health Study a larger leg length was inversely related to diabetes [8]. However, this association completely diminished when adjusted for BMI. Given the negative correlation between height and body fat percentage [24, 25], adjusting for BMI may at least in part inappropriately account for beneficial effects of a higher proportion of lean body mass with larger stature. Waist circumference reflects abdominal fat accumulation and was previously found to more strongly predict diabetes risk in our cohort [1]. In our study, associations of height with diabetes risk were observed in models adjusted for waist circumference. Inverse associations were not observable for sitting height in the Shanghai Women’s Health Study [8]. In the Atherosclerosis Risk in Communities study, leg length was inversely associated with diabetes risk [9]; however, models were not adjusted for BMI or waist circumference. In this study, the inverse associations became stronger after adjustment for body weight at age 25; however, such an adjustment likely makes interpretation of height components more difficult as they may become surrogates for lean body mass. Higher leg length/height ratio was inversely associated with risk in the Atherosclerosis Risk in Communities study for both white men and women, while we observed an inverse association more clearly for men.
Our data indicate a sex difference in associations of leg length vs sitting height: a larger sitting height at the cost of leg length (sitting height adjusted for total height) was related to increased risk in men, while among women both leg length and sitting height contributed to lower diabetes risk, although the latter association was statistically non-significant. This suggests that, among boys, growth before puberty, which relates more strongly to leg length, has a more favourable impact on later diabetes risk than growth during puberty (assuming that truncal bones are last to stop growing [26]), while for girls both growth periods seem to be important. However, our observation that sitting height (not adjusting for total height) is associated with lower diabetes risk is only partly in line with other studies [8, 10, 11, 27], which makes it difficult to conclude on sex differences in relation to different growth periods from our data. While sitting height has been related to lower insulin secretory function and insulin sensitivity [10], our results indicate that a detrimental effect among men can only be expected if growth of the trunk does not result in larger total height, thus being at the cost of leg length. The positive correlation between leg length and sitting height, however, suggests that such a growth pattern might not be common in our population.
Of note, waist circumference reflects abdominal fat accumulation, but also captures general features of skeletal structure and body size and thus scales to height [28]. We addressed this point in sensitivity analyses where models were adjusted for differences in waist circumference that would not be predicted based on height alone. Height remained inversely associated with diabetes risk in these analyses, although slightly more weakly compared with models with adjustment for measured waist circumference. Also, the inverse association between height and diabetes risk was more prominent among individuals with a BMI <25 kg/m2. This may indicate that a higher diabetes risk with larger waist circumference counteracts beneficial effects related to height, irrespective of whether larger waist circumference is due to growth or due to an energy imbalance.
Our results furthermore indicate that a substantial proportion of the inverse association between height and diabetes risk is attributable to lower liver fat content. Specifically, in women, adjustment for the FLI substantially weakened the associations between height and diabetes risk. That taller people have lower liver fat content has been described before [4], although data on this association are rather sparse. Ectopic lipid storage strongly affects the extent of insulin sensitivity [29] and may thus be a key characteristic explaining the link between greater height and lower diabetes risk. We evaluated a variety of cardiometabolic risk factors as potential explanations of the lower diabetes risk observed with greater height. Generally, BP, triacylglycerols, CRP and adiponectin appeared to be more strongly correlated with height among women and seemed to play a stronger role in the association of height with diabetes than among men. However, the tendency towards sex differences requires confirmation. Genetic studies have not identified sexually dimorphic associations with height [30], which speaks against sex differences in potential mediators. Still, specifically blood lipids (triacylglycerol and HDL-cholesterol) consistently attenuated associations between height and diabetes risk if adjusted for. These findings are supported by studies linking genetically determined height to cardiovascular disease and several cardiometabolic risk factors, including BP, triacylglycerol and CRP levels [13, 14]. However, the role of adiponectin in this context remains unclear. Adiponectin levels rather decrease than increase during growth among children small for gestational age [31, 32]. Also, only a few genetic variants relate to both adiponectin levels [33] and height, e.g. in LYPLAL1 and PDE3A. Still, it is conceivable that several overlapping and complex biologic pathways on the one hand influence height and on the other hand influence the risk of type 2 diabetes through an effect on lipid metabolism and function of adipose tissue (electronic supplementary material [ESM] Fig. 1). That these processes link height to the two key mechanisms that characterise type 2 diabetes (impaired beta cell function and insulin resistance) is supported by several studies. Tall people tend to have lower insulin resistance compared with shorter people [4, 11, 12, 24]. Growth during puberty and larger adult height have also been associated with concentrations of insulin-like growth factors [34, 35], which contribute to insulin sensitivity [36]. On the other hand, data from the Metabolic Syndrome in Men (METSIM) cohort suggest that height is also positively associated with beta cell function (disposition index) [12]. Interestingly, in this study greater height was related to improvements in both insulin sensitivity and beta cell function over time, independent of the baseline status of these two variables, age, waist circumference, physical activity and smoking. Still, whether the use of indices of insulin sensitivity and beta cell function derived from oral glucose tolerance tests in these studies is meaningful in the context of evaluation of height is questionable given that response to a fixed glucose load depends on the total amount of tissue for uptake and metabolism of glucose [37, 38]. It would be valuable to confirm such associations with gold standard methods of insulin sensitivity and secretion which account for differences in body size [39]. In our study, adjustment for HbA1c attenuated the association of relative leg length with diabetes in men. However, such adjustment could be considered an overadjustment given that HbA1c is used as a variable for diagnosing diabetes, although not at the time of our study. Although observational studies support an association between height and diabetes, investigation of genetically determined height has only suggested a trend for decreased risk of diabetes so far [13].
Our findings suggest that short people might present with higher cardiometabolic risk factor levels and have higher diabetes risk compared with tall people. Height can be used in diabetes risk prediction models, besides other risk factors. For example, in the German Diabetes Risk Score, points assigned per 1 cm of waist circumference or 1 year of age correspond to about 3 cm and 2 cm of height, respectively [6]. Thus, healthcare providers should be encouraged to consider height for risk assessment. On the other hand, attained height might represent an estimate of early childhood factors and their effects on later cardiometabolic risk. Thus, in terms of prevention of height-related diabetes risk, interventions likely need to focus on determinants of growth during pregnancy, early childhood, puberty and early adulthood. Although increased height was associated with reduced risk of type 2 diabetes, our data support that tallness is unlikely to modulate risk directly, but rather liver fat and other cardiometabolic risk factors are important mediators. Still, to what extent unfavourable risk factor profiles among shorter people require specific interventions remains unclear as height has not been systematically assessed as a modifying factor in this context.
The strengths of our study are that it was based on a large prospective cohort with measurements of components of stature (sitting height) and of a large variety of cardiometabolic risk factors that might explain associations of height with diabetes risk. Our study has limitations. For instance, it included only middle-aged men and women. Fat accumulation in the liver was estimated by the FLI. Although results should be replicated using more precise methods of liver fat determination, the FLI has been shown to correlate moderately with hepatocellular lipid content determined by magnetic resonance spectroscopy [40]. We have used the FLI as a proxy of liver fat content due to the absence of more precise measurements in the cohort. The FLI was strongly associated with risk of type 2 diabetes in comprehensively adjusted models (including adjustment for waist circumference) in the EPIC-Potsdam study [41]. Thus, our observations support the hypothesis that the lower risk in taller people is partially mediated by lower liver fat content. However, the adjusted model included waist circumference, an important component of the FLI, which might have led to an underestimation of the contribution of liver fat in our study. Most of the blood samples were collected non-fasting, which might have influenced specifically the interpretation of blood lipids and also the FLI in our study, which contains triacylglycerols in its calculation. However, when restricting the analysis to fasted participants, height remained inversely correlated with FLI in men (r = −0.22) and women (−0.15). Also, we considered only cardiometabolic markers as mediating factors being upstream from the two main pathophysiological mechanisms (insulin resistance and impaired beta cell function). Furthermore, as we do not have longitudinal data on height for our study participants, we were unable to account for height decline with ageing. Although such changes should have been small within our follow-up time [42], we cannot rule out that shorter height at study baseline has to some extent been the result of height loss prior to our study. Our analyses have been adjusted for age which should account both for age-related decline in height as well as age-cohort effects. Still, height loss has been observed to relate to higher risk of cardiovascular disease [43], but it remains unknown whether interventions to prevent age-related height decline would actually reduce diabetes risk. As for all observational studies, residual confounding might explain associations between height and diabetes risk. Specifically, parental socioeconomic status may relate to nutritional status during periods of growth and may insufficiently be reflected by participant education in our study. However, adjustment for parental socioeconomic status had little effect on associations between height and diabetes risk in the Atherosclerosis Risk in Communities study [9], and previous studies on genetic determinants of height support a biological interconnection between growth and cardiometabolic risk factors such as BP and lipid metabolism [13, 14].
In conclusion, we found an inverse association between height and risk of type 2 diabetes among men and women, which was largely related to leg length as a component of total height among men. Part of this inverse association may be driven by the associations of greater height with lower liver fat content and a more favourable profile of cardiometabolic risk factors, specifically, blood lipids, adiponectin and CRP.