In probing the directionality of the relationship between depression and obesity reported in cross-sectional studies, researchers have tested the hypothesis that the presence of obesity increases an individual's risk for depressive illness. Although studies of this question have been few, results from adult samples have reported a positive association.[22, 23] In adolescent samples, the longitudinal research is limited to four prospective studies, summarized in Table 1, which provide conflicting results. In a sample of rural, Caucasian participants of the Great Smoky Mountains Study who were followed from 9 to 16 years of age, Mustillo et al. found that chronically obese males were at greater risk of depression at 8-year follow-up than non-obese males. The same association was not found for females in the study. Conversely, Boutelle et al. reported an increased risk of depressive symptoms for obese adolescent females, followed for a 3-year period to a mean age of 16 years. In contrast, two studies of large, representative community samples that were followed for at least 15 years into adulthood demonstrated greater agreement in their findings with respect to females: both Herva et al. and Anderson et al. found adolescent obesity to be associated with the adult development of depression. In a sample from the longitudinal Northern Finland 1966 Birth Cohort Study, Herva et al. found that high BMI at age 14 years was associated with depressive symptoms at age 31 years for both males (OR 1.97, 95% CI 1.06–3.68) and females (OR 1.64, 95% CI 1.16–2.32). In comparison, Anderson et al. assessed adolescent participants of the upstate New York Children in the Community study at three time points, over a 20-year period, and showed that that the risk of depression was present for obese adolescent females only (HR 3.9, 95% CI 1.3–11.8). The different findings with respect to sex may be due in part to the different methods used to assess depression status in these studies: Herva et al. assessed depressive symptoms using a standardized self-report measure as compared with the structured diagnostic interview used by Anderson et al. In using a diagnostic interview to make depression diagnoses, Anderson et al. may have obtained a more accurate estimation of psychiatric illness. However, men in the Anderson study may also have under-reported depressive symptoms, as research suggests that males may be less likely to endorse depressive symptoms in interviews as compared with self-report measures, depending on the sex of the interviewer. As a result, an under-identification of depression may have occurred, leading to a lower likelihood of detecting an association between adolescent obesity and adult depression among men. In addition, by defining the outcome to symptoms at or above the diagnostic threshold for major depression, the study by Anderson et al. may have limited the relationship to that of more severe depressive pathology compared to the outcome of depressive symptoms, as measured by Herva et al. The discrepant findings by sex may also be due, in part, to differences in the studies' ability to retain participants over these long periods of time. Herva et al. achieved a 74% rate of follow-up, in contrast to a 96% rate of follow-up in the study by Anderson et al. As Herva et al. did not report demographic data for the missing individuals, it is not possible to make inferences about this group compared with those that continued in the study. It may be, for example, that non-depressed males were more engaged in their daily activities at the time of the follow-up survey and less likely to participate in the ongoing study assessment, thereby overestimating the association of early overweight with later depression in this study. Despite these caveats, however, the evidence to date suggests that while it is unclear if obese males are at greater risk for future depression, obese females are more likely to experience depression in adulthood than their normal-weight peers.