In this study, we sought to examine longitudinal associations of hostility measured in adulthood with BMI assessed at 4 points over a 19-year follow-up period. In general terms, there was clear evidence of a trend of increasing mean BMI over time in both sexes. Higher levels of hostility were associated with higher mean BMI at the start of follow-up in both men and women. Furthermore, results from analysis of the interaction with time showed that in men, the highest hostility levels were associated with increasing BMI during the 19-year follow-up period when compared with the lowest hostility levels. In women, the association between hostility and BMI remained constant over time. This implies that the effects of hostility on BMI in men and women track over time, with an increasing effect on BMI over time among men with the highest levels of hostility.
The present findings are in line with some previous studies showing hostility to be associated with higher BMI (3
). However, to the best of our knowledge, this was the first large-scale study to examine the longitudinal relation of hostility to BMI assessed repeatedly over an extended follow-up period. Since BMI has been shown to change considerably over time (18
), it is crucial to examine the dynamics of the association between hostility and BMI over time. The longitudinal modeling approach using time effects allowed us to control for weight gain over time, as well as to obtain precise estimates of effects. In contrast to some prior studies (3
), BMI in the present study was derived from height and weight assessed at a medical examination, thus minimizing measurement error or information bias and excluding the possibility of common-method bias. We were also able to control for self-reported BMI at age 25 years, which allowed us to examine the BMI trajectory over the adult life course.
As in other studies, BMI in the present study increased over time (18
). The result showing the highest hostility level to be associated with increasing mean BMI during follow-up in men is consistent with our hypothesis that coronary heart disease behavioral risk factors associated with hostility do not remain constant throughout a person's life. The fact that high hostility was associated with higher BMI throughout the follow-up period and also, in men, with an excess annual increase in BMI during follow-up lends support to the hypothesis that hostility as an individual personality characteristic may influence the development and maintenance of behavior-related risk factors (6
), evident in measures such as BMI.
There are several possible explanations for the link between hostility and BMI. First, the general cynicism and mistrust which characterize hostile persons may discourage them from following health promotion recommendations (35
). Cynicism may decrease the perceived importance of health-enhancing behaviors such as diet and physical activity (14
)—factors which have been found to be associated with greater BMI and obesity (22
). Second, lower socioeconomic status is associated with higher hostility (12
) and greater BMI (36
) and may be driving the association between hostility and BMI. However, in our analysis, this association remained robust to adjustment for socioeconomic status, either on its own or simultaneously with the other covariates. We assessed socioeconomic status using employment grade, the main measure of socioeconomic status in the Whitehall II Study. People in different grades differ with respect to salary, social status, and level of responsibility. Further research using repeated measures of socioeconomic status will be required to examine whether changes in socioeconomic circumstances mediate the association between hostility and BMI. An alternative explanation for this association is related to the psychosocial vulnerability model of hostility (14
). According to this model, hostile persons, given their oppositional attitudes and behaviors, are more likely to have increased interpersonal conflicts, lower levels of social support, more stressful life events, and a higher risk of depression (14
). The interrelations between these variables may influence BMI. For example, depression could result from a lack of social support or stressful life events and then affect diet or physical activity levels, which could in turn lead to higher BMI (38
). Here again, further research using repeated measures of depression is needed to examine this possibility.
In women, we observed increases in mean BMI during the follow-up period, but the association between hostility and BMI remained constant throughout follow-up. In other words, the term for interaction between time and hostility did not suggest that the association between hostility and BMI observed at baseline increased or decreased over time, even though it tracked over time. These sex-specific results suggest that the influence of hostility on BMI may be patterned by sex, perhaps because of sex-specific biologic phenomena. Coronary heart disease affects men more than women (39
), and it is not surprising that hostility, a risk factor for coronary heart disease (11
), is associated with an increasing effect on BMI over time in men alone. Menopause could be a confounder, as it has been found that at this time or several years before, women experience weight gain or have difficulty maintaining their usual weight (40
In interpreting the present results, it is important to note 2 limitations. First, our cohort of civil servants did not include blue-collar workers and unemployed people and thus is not representative of the general population, which may limit the generalizability of our findings. Second, we were able to control for BMI at age 25 years, but this variable was derived from self-reported height and weight. However, in our results, BMI at age 25 years was found to be strongly associated with objectively measured BMI in phase 1, which supports the validity of this measure.
In summary, we found that mean BMI increased over a 19-year follow-up period among both men and women in the Whitehall II Study. We also found prospective evidence for an effect of hostility on BMI over the 19-year follow-up period. Finally, higher hostility was associated with significantly greater increases in BMI over time in men, suggesting that difference in BMI as a function of hostility is not stable over time. These results have implications for studies (e.g., studies on the association between hostility and coronary heart disease) in which the association between hostility and health outcomes is adjusted for health behaviors like BMI at baseline in order to assess the “independent” effect of hostility on health. Our findings suggest that controlling for baseline BMI might not be sufficient to address the mediation effect, particularly in men. Epidemiologic studies with repeated measures of covariates are now widespread. Going beyond baseline covariates might allow proper modeling of the mechanisms underlying the association between hostility and important health outcomes, such as coronary heart disease.