What is the impact of underweight on self-reported health trajectories and mortality rates: a cohort study.

BACKGROUND: Utilizing a cohort study design combining a survey approach with repeated physical examinations, we examined the independent effects of BMI on mortality and self-reported health (SRH) and whether these independent effects change as people grow older.

METHODS: The Tromsø Study consists of six surveys conducted in the municipality of Tromsø, Norway, with large representative samples of a general population. In total, 31,985 subjects participated in at least one of the four surveys administered between 1986 and 2008. Outcomes of interest were SRH and all-cause mortality.

RESULTS: Overweight and underweight subjects reported significantly lower levels of SRH, but age affected the thinnest subjects more than all others. The SRH trajectory of underweight subjects at age 25 was slightly above the other categories (0.08), but it fell to -.30 below the reference category at age 90. For obese subjects, the difference was -0.15 below the reference category at age 25 and -0.18 below at age 90. This implies that even though a low BMI was slightly beneficial at a young age, it represented an increasing risk with age that crossed the reference curve at age 38 and even crossed the obese trajectory at age 67 in the full fitted model. The proportional hazard ratio for those who were underweight was 1.69 (95% CI: 1.38-2.06) for all-cause death as compared to 1.12 (95% CI: 1.02-1.23) for obese subjects.

CONCLUSION: BMI affected SRH and all-cause mortality independently from comorbidity, mental health, health-related behaviors and other biological risk factors. Being underweight was associated with excess mortality as compared to all others, and age affected the thinnest subjects more than all others. Weight increase was beneficial for mortality but not for SRH among the underweight. The rapid decline of SRH with increasing age suggests that particular attention should be paid to underweight after 38 years of age.