In this randomized trial comparing diets with modified CHO content with a clinically standard portion-controlled diet (effectively an energy-reduced, low-fat diet), the central finding was that all 3 diets were efficacious in improving BMI z-score for up to 9 months post-intervention. These findings are consistent with other diet comparisons in obese adolescents16
and adults.9, 12, 13, 32, 33
However; the results of our study do not support the hypothesis that CHO-modified diets would be more effective in improving weight status than a standard portion-controlled diet among obese children. Subjects in all 3 diet groups showed significant improvements in adiposity measures (BMI z-score, %BF and WC) during the initial 3-month intervention with relatively intensive intervention and contact with study dietitians. At 12-months, improvements in BMI z-score and %BF were maintained in all 3 diet groups; however, none of the groups sustained improvement in WC. These results are similar to clinical trials of the LC and low-fat diets in a 12-week study of adolescents16
and a 12-month study of adults, 12
reporting comparable improvements in weight status. However, other clinical trials found the LC diet to be superior to a low-fat diet with both adolescents15
and adults.8-11, 13,14
In addition, controlled studies of the RGL diet with obese children and adolescents showed a greater decrease in BMI17, 18
and body fat mass18
than those assigned to low-fat diet regimens. Factors behind these discrepant results are not clear at this time.
It is noteworthy that subjects in all diet groups were successful in maintaining a reduced caloric intake and altered dietary composition even in the final 9 months without intervention contact. These findings raise the possibility that intensive guidance with the initial clinical application of weight management diets could lead to long-term success in children.
Each of the diets led to improvement in some of the clinical measures, but there was variability among the results. The LC group had significant improvement in HDL cholesterol and triglycerides, and the PC and RGL diets seemed to have a relatively greater effect on lowering fasting insulin and glucose. Other clinical trials of the LC and low-fat diets also reported improvements in metabolic parameters.8-11, 15, 16
It is not clear the degree to which the diet’s macronutrient composition in our study contributed to these variable outcomes because they are undoubtedly also affected by other factors. Moreover the clinical implications of these findings are not clear given the relatively normal baseline values. Nonetheless, none of the diets had systematic adverse effects on incidence of cardiovascular risk factors, and were in this regard safe.
Recently it has been suggested that strategies to individualize diet interventions to match patients’ metabolic profiles could have increased efficacy.34, 35
For example, a carbohydrate-modifying diet may be advantageous in subjects with elevated triglycerides, and a low-fat diet may be more effective in individuals with high LDL cholesterol.36
Our trial provides limited support for these hypotheses, as we detected differential effects on specific metabolic profiles (e.g., lipids vs. insulin resistance) by dietary intervention. More trials matching dietary approaches with children’s specific risk factors are greatly needed.
The LC diet led to notable long-term improvements in both anthropometric and clinical metabolic profiles, despite several indications that it was a less acceptable diet to participants. The LC diet was the only group to which individuals refused randomization, and randomized participants had significantly lower dietary adherence compared with the RGL and PC groups. A question raised by the study’s findings is whether adherence to a less stringent version of the LC diet is more realistic, sustainable, and yet still effective for improving BMI z-score and other obesity-related clinical measures. The study results showed strict adherence to the LC diet was extremely difficult to achieve with children, (i.e., very few Keto-Diastix® results were positive for ketones in the urine). However, the reported mean carbohydrate intake in the LC group was still significantly less than in the RGL and PC groups at all time points, and resulted in significant improvements in BMI z-score and some clinical measures after one year. Although not statistically significant, this study also suggests that the LC group’s longer-term results may be associated with adherence. For example, a trend toward rebound in %BF and BMI z-score occurred when dietary adherence dropped to near zero in this group. Larger scale studies would be required to assess this relationship.
By contrast, the sustained high adherence to the RGL diet coupled with anthropometric and clinical improvements may make it the most promising intervention for the long-term weight management of children. Alternatively, prescribing one of the 3 diets based on the patient’s preferences for the initial intervention and then transitioning to the RGL diet for longer-term maintenance may also be an effective strategy. A prior retrospective study showed elevated fasting insulin in obese youth is a risk factor for being unsuccessful with weight management (i.e. increased BMI z-score) when participating in an obesity intervention that included a RGL diet.37
However our study does not reflect this finding. The conflicting results may be explained in part by the difference in study design as well as the level of engagement and motivation of our study population as evidenced by a high retention rate.
In this randomized clinical trial with obese children, we tested the long-term efficacy and safety of 3 diet interventions using similar evidence-based behavioral strategies. The intervention was delivered in a consistent manner with strong family-based behavioral components across the 3 diet groups with quality control procedures to ensure treatment fidelity. The strong retention rate (≥ 77% of subjects) increased the validity of reported findings, and gave the study adequate power despite enrollment levels less than the initial target of 50 subjects per group. We note that the PC group included a lower proportion of boys than the other groups; although this was not statistically significant, we may have lacked power to detect true differences. However, we believe this would have a minimal impact on the conclusions of this study, given subjects were stratified according to sexual maturity and extent of overweight.
It is unknown whether the reported improvements would be sustained beyond one year or whether one diet would be more effective than the others in the longer term. Because the study was implemented in a research context and the most severely obese children were excluded, the ability to generalize the findings to a clinical population of obese children is unknown. Lastly, the reliance on self-report of dietary intake is a limitation that is encountered by most diet trials. Underreporting of dietary intake for obese adults is well documented. Thus it seems likely that this phenomenon would have occurred with equal magnitude among the 3 diet groups.
We conclude that the three diets that differed in macronutrient content and glycemic load resulted in similar and significant improvement in BMI z-score and related health measures in children. Though strict adherence to the LC diet was more difficult to achieve in children, all diets were effective in improving some adiposity and clinical outcomes. These findings suggest that practitioners may offer any of these dietary approaches for achieving a healthier weight with obese children.