Weight development in children with obesity without treatment: A Danish cohort study with long-term follow-up

Abstract

Introduction: Limited insight exists into the weight development in children with obesity not receiving obesity treatment.

Methods: This cohort study included 467 Danish children aged 5-10 years with obesity (iso-BMI >30 kg/m²) not receiving treatment. Data from mandatory health check-ups on school-children's height and weight (converted to BMI z-scores) were merged with the Danish National Registries. A multivariable logistic regression weighted for the duration of follow-up was used to estimate odds ratios (OR) for normalization of BMI (iso-BMI 18.5-25 kg/m²) and obesity remission (iso-BMI 18.5-30 kg/m²).

Results: During a median follow-up of more than 6 years, 7.9% of the children normalized their BMI, while 45.4% obtained obesity remission. BMI z-score at inclusion acted as a strong inverse predictor for normalizing BMI (OR 0.14 per one-unit SD, CI: 0.03-0.53) and for obesity remission (OR 0.17 per one-unit SD, CI: 0.08-0.37). No other significant predictors were observed in the weighted multivariable models.

Conclusion: Higher BMI z-scores inversely predict normalizing BMI and achieving obesity remission in untreated children. Given that many children naturally achieve obesity remission or weight normalization, resources should focus on understanding barriers of obesity maintenance and to develop effective strategies for those who do not experience improvement.

FIGURE 1

Flowcharts for inclusion and exclusion of children with obesity aged 5–10 living followed for a minimum of 3 years living in Aarhus Municipality, Denmark between 1 January 2010 and 17 March 2018, and not invited to participate in the lifestyle intervention.

FIGURE 2

Logistic regression models, including 467 children aged 5–10 years living with obesity and a minimum of 3 years of follow‐up not receiving a lifestyle intervention, present the OR of normalizing weight at the end of follow‐up with having overweight or persistent obesity as the reference group. For each co‐variable, we present univariable, unweighted multivariable, and weighted multivariable OR. The included co‐variables are sex (boys as reference), BMI z‐score (per unit increase in SD), and age (per unit increase in a year). The weighted multivariable model was weighted by the ratio of follow‐up of each individual relative to the maximum follow‐up observed in the cohort.

FIGURE 3

Logistic regression models, including 467 children aged 5–10 years living with obesity and a minimum of 3 years of follow‐up not receiving a lifestyle intervention, and presenting the odds ratio of being in the obesity remission group (212 children) at the end of follow‐up with the persistence of obesity group as the reference. For each co‐variable, we present univariable, unweighted multivariable, and weighted multivariable odds ratios. The included co‐variables are sex (boys as reference), BMI z‐score (per unit increase in SD), age (per unit increase in a year) highest completed household education (low as reference), family type (two‐adult family as reference), immigration status (Danish origin as reference), psychiatric diagnosis at the child (no diagnosis as reference) or family history of mental illness (no disposition as reference). The weighted multivariable model was weighted by the ratio of follow‐up of each individual relative to the maximum follow‐up observed in the cohort.

FIGURE 4

Logistic regression models, including 467 children aged 5–10 years living with obesity and a minimum of 3 years of follow‐up not receiving a lifestyle intervention, and presenting the odds ratio of being above 80th percentile for change in BMI z‐score with being below the 80th percentile for change in BMI z‐score (375 children) as reference. For each co‐variable, we present an univariable, unweighted multivariable, and weighted multivariable odds ratio. The included co‐variables are sex (boys as reference), BMI z‐score (per unit increase in SD), age (per unit increase in a year) highest completed household education (low as reference), family type (two‐adult family as reference), immigration status (Danish origin as reference), psychiatric diagnosis at the child (no diagnosis as reference) or family history of mental illness (no disposition as reference). The weighted multivariable model was weighted by the ratio of follow‐up of each individual relative to the maximum follow‐up observed in the cohort.