Weight loss intervention may not benefit all patients with type 2 diabetes

Liz Meszaros, MDLinx | July 19, 2017

Patients with type 2 diabetes with well-controlled diabetes but poor self-reported general health may not derive health benefits from intensive weight loss interventions, and clinicians may be better served screening for HbA1c and applying a short questionnaire on general health to identify these patients before beginning such interventions, according to a recent analysis of data from the Action for Health in Diabetes (Look AHEAD) trial, published in The Lancet Diabetes & Endocrinology.

Advertisement

Does intensive weight loss intervention work in T2D?

T2D patients with well-controlled diabetes and poor self-reported general health may not benefit.

“This analysis restores my faith in basic common sense,” said co-author Ronald Tamler, MD, medical director, Mount Sinai Clinical Diabetes Institute. “For the vast majority of people with diabetes, a healthy lifestyle with weight loss carries significant benefits; however, it’s not for everyone. Thanks to this work, clinicians can infer which patients will benefit the most from such a lifestyle intervention.”

The Look Ahead trial was halted in 2012 because results were not statistically significant, but Dr. Tamler and fellow researchers at The Arnhold Insitute for Global Health of the Icahn School of Medicine at Mount Sinai applied new, advanced machine learning techniques to these data. Using these techniques, they sought to assess whether weight loss interventions in people with type 2 diabetes could bring about reductions in long-term cardiovascular disease morbidity and mortality.

They included 4,901 of the original 5,145 patients with type 2 diabetes enrolled in the Look AHEAD randomized, controlled study. For model development, they included 2,450 patients, and for the testing dataset, 2,451.

Using causal forest modeling, which identified heterogeneous treatment effects (HTEs) from intensive weight loss interventions, researchers used a random half of the trial data (the training set), and then applied Cox proportional hazards models to assess the potential HTEs on the remaining 50% of the data (the testing set).

Cox modeling for the primary composite cardiovascular outcomes showed the number needed to treat to prevent one event over 9.6 years was 28.9 in subjects with an HbA1c of 6.8% or greater, or both an HbA1C less than 6.8% and a Short Form Health Survey (SF-36) general health score of 48 or more (2,101 of 2,451 subjects in the testing dataset; 167 of 1,046 primary outcomes events for intervention vs 205 of 1,055 for control) (absolute risk reduction: 3.46%; 95% CI: 0.21-6.73%; P=0038).

In contrast, subjects with an HbA1c of less than 6.8% and baseline SF-36 general health score of less than 48 (350 of 2,451 subjects in the testing data; 27 or 171 primary outcome events for intervention vs 15 of 179 primary outcome events for control) had an increase in absolute risk for the primary outcome of 7.41% (95% CI: 0.60 to 14.22; P=0.003).

Thus, cardiovascular events were prevented in subjects with moderately or poorly controlled diabetes (HbA1c 6.8% or higher) and those with well-controlled diabetes (HbA1c less than 6.8%) and good self-reported health.

In 15% of subjects with well-controlled diabetes and poor self-reported general health, however, researchers observed negative effects, which rendered the overall study outcome neutral. In this subgroup of patients, researchers also found substantially poorer compliance with the exercise portion of the intervention, and less improvement in several intermediate health outcomes, including glucose levels, mental health, and blood pressure.

“Our analysis demonstrates that recent advances in machine learning for causal inference can increase the quantity of clinically relevant findings generated from large randomized trials,” said lead author Aaron Baum, PhD, lead economist, The Arnhold Institute for Global Health; assistant professor, department of health system design and global health, Icahn School of Medicine at Mount Sinai. “As researchers and data scientists, we are always concerned that an overall study result could mask important disparities in benefit or harm among different types of patients, which is exactly what this study revealed. Being able to identify individuals who could benefit from an intervention is fundamental to patient care.”

Prabhjot Singh, MD, PhD, director of The Arnhold Institute for Global Health and Chair of the Department of Health System Design and Global Health, concurs with Dr. Baum: “This research strengthens the role for data science and precision medicine as essential tools that can transform the way health care is delivered. Identifying individuals who could benefit from an intervention is crucial for practicing clinicians, while ignoring subgroups that benefit might lead to lack of reimbursement for weight loss programs, which would neglect vulnerable populations.”

Advertisement