The association between adult attained height and sitting height with mortality in the European Prospective Investigation into Cancer and Nutrition (EPIC)

Adult height and sitting height may reflect genetic and environmental factors, including early life nutrition, physical and social environments. Previous studies have reported divergent associations for height and chronic disease mortality, with positive associations observed for cancer mortality but inverse associations for circulatory disease mortality. Sitting height might be more strongly associated with insulin resistance; however, data on sitting height and mortality is sparse. Using the European Prospective Investigation into Cancer and Nutrition study, a prospective cohort of 409,748 individuals, we examined adult height and sitting height in relation to all-cause and cause-specific mortality. Height was measured in the majority of participants; sitting height was measured in ~253,000 participants. During an average of 12.5 years of follow-up, 29,810 deaths (11,931 from cancer and 7,346 from circulatory disease) were identified. Hazard ratios (HR) with 95% confidence intervals (CI) for death were calculated using multivariable Cox regression within quintiles of height. Height was positively associated with cancer mortality (men: HRQ5 vs. Q1 = 1.11, 95%CI = 1.00–1.24; women: HRQ5 vs. Q1 = 1.17, 95%CI = 1.07–1.28). In contrast, height was inversely associated with circulatory disease mortality (men: HRQ5 vs. Q1 = 0.63, 95%CI = 0.56–0.71; women: HRQ5 vs. Q1 = 0.81, 95%CI = 0.70–0.93). Although sitting height was not associated with cancer mortality, it was inversely associated with circulatory disease (men: HRQ5 vs. Q1 = 0.64, 95%CI = 0.55–0.75; women: HRQ5 vs. Q1 = 0.60, 95%CI = 0.49–0.74) and respiratory disease mortality (men: HRQ5 vs. Q1 = 0.45, 95%CI = 0.28–0.71; women: HRQ5 vs. Q1 = 0.60, 95%CI = 0.40–0.89). We observed opposing effects of height on cancer and circulatory disease mortality. Sitting height was inversely associated with circulatory disease and respiratory disease mortality.


Introduction
Poor nutrition, illness and early life exposures may contribute to ill health in later life [1][2][3]; however, there is a paucity of data to explore such associations in prospective cohorts with extended follow-up of children. Adult height is an easily measured variable, and is thought to reflect both genetic and environmental factors including nutrition, physical and social environments in early life [4,5].
The association between height and mortality has been investigated in previous studies. A meta-analysis of 121 cohort studies comprising over 1 million participants reported that height was inversely associated with risk of death from circulatory diseases such as coronary disease, stroke and heart failure [6]. In contrast, height was positively associated with risk of death from melanoma and cancers of the pancreas, endocrine and nervous systems, ovary, breast, prostate, colorectum, blood and lung [6]. Despite many studies investigating overall height and mortality, there have been few studies examining the association between sitting height and mortality. Higher sitting height is of interest because compared with adult height, sitting height may be more strongly positively associated with insulin resistance [7], and is positively associated with lung function, independently of height [8]; therefore, the effects of sitting height on mortality might be different from those of overall height. One cohort study reported that sitting height was positively associated with cancer mortality and inversely associated with death from circulatory disease [9], but others showed no association [7,[10][11][12].
To further knowledge on the association of height and health outcomes among adults, we examined whether adult height and sitting height were associated with overall and cause-specific mortality in a large prospective cohort of approximately half a million men and women from 10 European countries.

Study cohort
The European Prospective Investigation into Cancer and Nutrition (EPIC) study includes 23 centres within 10 European countries (Denmark, France, Germany, Greece, Italy, the Netherlands, Norway, Spain, Sweden, and the United Kingdom (UK)). Most centres recruited from the general population living in defined towns and provinces. In Florence (Italy) and Utrecht (the Netherlands), however, participants undergoing breast cancer screening were recruited; parts of the Italian and Spanish cohorts were recruited among blood donors and their spouses; most of the Oxford cohort (UK) consisted of vegetarian and health-conscious volunteers; and female members of the health insurance scheme for state school employees were recruited in France. Between 1992 and 2000, 521,457 individuals (approximately 70% women, mostly 20-70 years old) were enrolled after providing written informed consent. Ethical approval for the EPIC study was obtained from the review boards of the International Agency for Research on Cancer (IARC) and local participating centres. The cohort characteristics have been described in detail elsewhere [13,14].

Exposure assessment
At recruitment, standardized questionnaires on lifestyle, demographic information and personal history were collected [13,14]. Height was measured in participating centres to the nearest 0.1, 0.5, or 1.0 cm in participants without shoes on [15]. Norway was the only centre in which height was all self-reported; furthermore, height was only measured in 29% of the French and 13% of the Oxford cohort with the remainder of the participants self-reporting their height. Self-reported data on height tends to be overestimated, with the degree of overestimation being larger for shorter individuals and this also depends on age [15]. The self-reported data from the Oxford cohort were adjusted using earlier described sex-specific regression equations that incorporated age [15]; this was not done for the French cohort because the interval between the self-reported data and measurement was considered to be too long to do so reliably, thus only those participants with measured data were included. Sitting height was measured (the minimum unit was 0.1cm) in over 90% of participants in six countries (Italy, Spain, the Netherlands, Germany, Greece, Denmark) and in 29% of French participants.

Follow-up and endpoint assessment
Vital status, causes and dates of death were ascertained from population registries in Denmark, Italy (except Naples), the Netherlands, Norway, Spain, Sweden, and the UK. In Germany, Greece and Naples, this information was obtained by follow-up mailings or inquiries to municipal registries, regional health departments, physicians, and hospitals, and also by directly contacting their next-of-kin. In France, the causes and dates of death were obtained from the French Epidemiological Center for the Medical Causes of Death (CépiDc, Inserm).

Statistical analysis
From the 521,457 participants recruited, those in a subsample in France (n = 52,809) and in the whole cohort in Norway (n = 37,185) were excluded because measured height was unavailable. Additionally, participants with missing questionnaire data (n = 1,286), missing dietary data (n = 6,627), missing all potential confounders (n = 3,127), without dates of death or follow-up information (n = 542) and those within the lowest and highest 1% of the cohort distribution of the ratio of reported total energy intake to energy requirements (n = 10,133) were excluded [17]. The final analytic cohort consisted of 409,748 individuals. For the analysis of sitting height, participants whose sitting height was not assessed were also excluded, leaving an analytic cohort for sitting height of 253,427 individuals.
Cox proportional hazard regression models with age as the underlying time scale were used to estimate hazard ratios (HR) and 95% confidence intervals (CI) for the association between height, sitting height and mortality risk. Height and sitting height were analysed as categorical variables defined by quintiles, and rounded off to the nearest 1 cm, and as continuous variables. Time at risk was estimated from the date of recruitment to the date of death, emigration, loss to follow-up, or the end of follow-up (a maximum through 2010, depending on centre), whichever occurred first. To control for differences in questionnaire design and follow-up procedures, all models were stratified by study centre. Models were further stratified by age at recruitment (continuous) to allow the form of the baseline hazard functions to vary across ages. All models were fitted for men and women separately, and adjusted for weight (kg, quintiles), combined recreational and household physical activity (inactive, moderately inactive, moderately active, active, unknown), alcohol consumption (0, >0-4, 5-14, 15-29, 30-59, ! 60 g/day), smoking status (never, former smokers (who quit <10 or !10 years ago), current smokers (1-14, 15-24 or !25 cigarettes per day), current smoker but amount missing (unknown smoking status), education level (none/primary school, technical or professional school, secondary school, university degree, not specified/missing) and energy intake (kcals/day, continuous). Moreover, the Cox models for women were further adjusted for menopausal status (pre, post-, peri-menopausal or unknown) and menopausal hormone use (yes, no, unknown). Potential non-linearity of the dose-response relationship was investigated using restricted cubic spline regression with knots placed at the 5 th , 25 th , 75 th and 95 th percentiles of height and corresponding likelihood ratio tests to compare the goodness-of-fit of the models with and without the spline terms [18,19]. Because linearity could indeed be assumed, we computed a test for trend based on models with height and sitting height as a continuous variable.
Finally, we conducted interaction analyses to examine the relation between sitting height and overall height in relation to mortality; for these analyses we categorized height and sitting height into tertiles (low, middle and high) and examined risks within each combined strata and calculated a P-value for the interaction term. In addition, we examined interaction terms for height (as a continuous variable, per 5cm increment) and all-cause and cause-specific mortality according to country, age at recruitment, body mass index (BMI), smoking status and alcohol intake.
All reported P-values were two-sided and were regarded as statistically significant if P<0.05. The potential for multiple comparisons was addressed by examining Bonferroni correction; p = 0.004 (0.05/13 variables). All analyses were performed using Statistical Analysis Software (SAS version 9.1, SAS Institute, Cary, NC).

Results
After an average follow-up of 12.5 years (range 0.01-17.8), 29,810 participants (15,320 men and 14,490 women) had died from any cause among all 409,748 participants. Out of all deaths with a reported cause (n = 25,526), major causes were cancer (n = 11,931), diseases of the circulatory system (n = 7,346) and the respiratory tract (n = 1,266). Among participants with data on sitting height, there were 15,630 all-cause deaths, 6,909 deaths from cancer, and 3,656 deaths from circulatory diseases. Participants who were taller, compared with those who were shorter, were younger, heavier, had higher energy intakes and were more physically active. The proportion of current smokers was lower in taller men but there was a higher proportion of current smokers in taller women; furthermore, taller women were more likely to have a higher education level, consume higher levels of alcohol, to be premenopausal, and among post-menopausal women, were more likely to use menopausal hormones (Table 1). After excluding individuals without data for sitting height, the characteristics were similar to those with measured height (data not shown). Tables 2 and 3 shows the HRs for height and all-cause and cause-specific mortality in men and women, respectively. There was a statistically significant linear inverse association between height and all-cause mortality in men (HR Q5 vs. Q1 = 0.85, 95%CI = 0.80-0.91, p for trend<0.01), but no association was observed in women (HR Q5 vs. Q1 = 1.01, 95%CI = 0.95-1.08, p for trend = 0.66). We observed a positive association between height and death from cancer in both sexes (HR Q5 vs. Q1 = 1.11, 95%CI = 1.00-1.24, p for trend = 0.08 in men; HR Q5 vs. Q1 = 1.17, 95%CI = 1.07-1.28, p for trend<0.01 in women). HRs for smoking-related cancers and non-smoking-related cancers were not substantially different (Table 3). In contrast, height was inversely associated with the risk of death from circulatory disease in both sexes (HR Q5 vs. Q1 = 0.63, 95%CI = 0.56-0.71, p for trend<0.01 in men; HR Q5 vs. Q1 = 0.81, 95%CI = 0.70-0.93, p for trend<0.01 in women). Furthermore, height was inversely associated with ischaemic heart disease and myocardial infarction in both men and women, as well as cerebrovascular disease mortality in men only. There was no association between height and death from stroke or respiratory diseases in men or women. Excluding subjects with a past history of cancer, cardiovascular disease or diabetes (n = 38,760) yielded similar results (data not shown). Tables 4 and 5 shows the HRs for sitting height and all-cause and cause-specific mortality in men and women, respectively. We observed inverse associations for sitting height and allcause mortality in both men and women (HR Q5 vs. Q1 = 0.81, 95%CI = 0.74-0.88, p for trend<0.01 in men; HR Q5 vs. Q1 = 0.86, 95%CI = 0.79-0.94, p for trend<0.01 in women). In contrast to the findings for overall height, we did not observe an association between sitting height and cancer mortality in either men or women. The associations between sitting height and circulatory disease mortality were similar to the inverse findings for overall height. In addition, we observed an inverse association between sitting height and death from haemorrhagic stroke in men only (HR Q5 vs. Q1 = 0.44, 95%CI = 0.23-0.84, p for trend<0.01) and from respiratory disease in men and women (HR Q5 vs. Q1 = 0.45, 95%CI = 0.28-0.71, p for trend <0.01; HR Q5 vs. Q1 = 0.60, 95%CI = 0.40-0.89, p for trend<0.01, respectively). When we analysed the association between sitting height and mortality, additional adjustment for overall height did not substantially change our results.
We also investigated interactions between sitting height and overall height in relation to mortality (Table 6). Taller overall height and taller sitting height was strongly inversely associated with death from respiratory disease in men (P interaction = 0.03) but not women. However, there were no significant interactions between sitting height and overall height in relation to all-cause, cancer or circulatory disease mortality. Interaction terms for overall height and   * Hazard ratios (HR) and 95% confidence intervals (95% CI) estimated in a Cox regression model stratified by centre,    * Hazard ratios (HR) and 95% confidence intervals (95% CI) estimated in a Cox regression model stratified by centre,  * Hazard ratios (HR) and 95% confidence intervals (95% CI) estimated in a Cox regression model stratified by centre, 1-year age and 5-year birth cohort categories, adjusted for education level (none,/primary school completed, technical/professional school, secondary school, university degree, not specified), smoking status (never smoker, former smoker who quit <10 years ago, former smoker who quit >10 years ago, former smoker, unknown when quit, current smoker of 1-14 cigarettes a day, current smoker of 15-24 cigarettes a day, current smoker of > = 25 cigarettes a day, current smoker but amount missing, smoking status unknown), physical activity (inactive, moderately inactive, moderately active, active), alcohol consumption (0, 0-<5, 5-<15, 15-<30, 30-<60, > = 60 g/day), weight (quintiles), energy intake (kcals/day, continuous), menopausal status (pre, post-, peri-, unknown) and menopausal hormone use (yes, no, unknown). † Median values of each category as continuous variable (cm). doi:10.1371/journal.pone.0173117.t005 mortality by age, smoking status (smoker/non-smoker), and alcohol intake (high/low) did not reveal meaningful differences; however, for BMI, we observed an increased risk per 5cm increment in height for all-cause mortality only among women with a BMI <25 kg/m 2 (data not shown). After Bonferroni correction, the significance of our results was not substantially changed.

Discussion
In this large prospective study, overall height was positively associated with deaths from cancer, but inversely associated with deaths from circulatory disease. These results are supported by a previous meta-analysis of 1 million people from 121 prospective studies [6]. In the present study, sitting height was not associated with cancer mortality but was inversely associated with all-cause mortality, circulatory deaths, and death from respiratory disease. To our knowledge, this is the first study to report an inverse association between sitting height and death from respiratory disease. The World Cancer Research Fund/ American Institute for Cancer Research (WCRF/AICR) reported that there is convincing data that height increases the risk of individuals being diagnosed with cancers of the colorectum, breast (postmenopausal) and ovary [20,21]; furthermore, height 'probably' increases the risk of cancers of the pancreas and breast (premenopausal). To complement this previous data on incidence, our data suggests a role for height and risk of cancer mortality. Short stature is a well-documented risk factor for mortality from circulatory diseases [6,9,22], ischemic heart disease [23], ischemic stroke [22,24,25] and haemorrhagic stroke [24,25] in previous studies. The results of this current analysis corroborate these prior studies but we also observed strong inverse associations between height and subtypes of circulatory disease death despite their different pathologies.
Whether a relationship between sitting height and mortality also exists is largely unknown. Wang et al. reported height and sitting height were positively associated with cancer death, but were inversely associated with death from cardiovascular disease in a cohort of 135,000 Chinese men and women [9]. Four other studies reported no association between sitting height and mortality [7,[10][11][12]. Our study in a large European population generally supports the reports from the Chinese, although we did not find a positive association between sitting height specifically and cancer mortality.
There are several potential underlying mechanisms to explain the opposing association of adult height with circulatory disease and cancer mortality. The positive association between adult height and mortality from cancer may be a result of taller people having larger organs, and a greater number of cells at risk of malignant transformation and/or proliferation [26]. Furthermore, attained adult height is known to be related to early nutrition in childhood or adolescence [3,5]. In contrast, the inverse association between height and cardiovascular disease mortality has been proposed to be due to taller people and people with higher sitting height having larger coronary vessel diameters and a slower heart rate and/or greater lung capacity [8,[27][28][29][30]. Height may also be a marker of early exposure to components of the insulin/growth hormone axis. Height is correlated with circulating levels of insulin-like growth factor (IGF)-I, the main mediator of growth hormone activity and a hormone that has been positively associated with cancers at a number of anatomic sites [31][32][33][34][35], but IGF-1 levels are generally inversely related with circulatory disease risk [36][37][38][39][40]. Crowe et al. reported that each 10cm increase in height corresponded to a 4% increase in circulating IGF-1 levels [41]; therefore, increasing IGF-1 levels might mediate the opposing effect of height on cancer and circulatory disease mortality. To clarify the underlying mechanisms, further studies are needed to investigate IGF-1 levels in relation to cause-specific mortality risk while accounting for adult height. Furthermore, several genetic factors are related with height, cancer and cardiovascular disease [42,43]. Identifying such genetic variants might shed light on potential mechanisms underlying the associations between height and mortality.
Davey Smith et al. reported that sitting height was strongly positively associated with insulin resistance [7]; thus, we expected a clearer association between sitting height and cancer mortality than overall height. Despite finding a positive association between overall height and cancer mortality in our data, there was no association for sitting height. These null findings for sitting height and cancer mortality may be plausible because despite the association with insulin resistance, sitting height has been associated with improved prognosis in cancer survivors due to better lung function in those with greater sitting height [8,44]. Our finding that sitting height was inversely associated with death from respiratory disease is of note as this may be due to the aforementioned association between sitting height and lung function [8,44].
Height is positively associated with education level among women in this study. Previous studies reported that lower educational levels have been associated with increased mortality, and incidence of coronary heart disease and stroke in Europe and the United States [45][46][47].
In an attempt to control for this potential confounder, we adjusted our models for educational level, although our results did not change from the unadjusted models.
A major strength of the EPIC study is the large study population representing findings from multiple countries and its long follow-up, resulting in a large number of deaths allowing us to analyse and distinguish between different causes of death. This study enabled us to examine measured height on the majority of participants, to adjust the self-reported height variable in the others, and to examine measured sitting height in a large subset of the cohort. In contrast, this study had some limitations. With a large body of information on lifestyle variables, we could adjust for many potential confounding factors, although the possibility of residual confounding cannot be excluded. Additionally, we divided cardiovascular disease into subgroups, which may result in some degree of misclassification.
In conclusion, this study revealed opposing findings for the relationship between height on cancer and circulatory disease mortality. Specifically, we showed that height was positively associated with death from cancer, but inversely associated with death from circulatory disease. Furthermore, this is the first study to show the inverse association between sitting height and death from respiratory disease. These findings could be used to contribute to risk prediction models to target individuals for specific screening programmes.