Heterogeneity of Treatment Effects From an Intensive Lifestyle Weight Loss Intervention on Cardiovascular Events in Patients With Type 2 Diabetes: Data From the Look AHEAD Trial

Literature - De Vries TI, Dorresteijn JAN, Van der Graaf Y et al., - Diabetes Care 2019; dc190776. https://doi.org/10.2337/dc19-0776

Introduction and methods

Weight loss beneficially impacts metabolic control and CV risk factors in patients with type 2 diabetes (T2DM) [1,2]. Bariatric surgery may lower the risk of CV events in obese patients with T2DM [3]. It is, however, unknown whether weight loss through lifestyle intervention positively impacts CV outcomes in patients with T2DM.

The randomized Action for Health in Diabetes (Look AHEAD) trial evaluated an intensive lifestyle intervention program that promoted weight loss through decreased caloric intake and increased physical activity in overweight and obese patients with T2DM. The control group received regular diabetes support and education [4]. After a median follow-up of 9.6 years, the Look AHEAD trial was stopped prematurely due to a futility analysis. The intervention was found to lead to greater weight loss and greater reductions in HbA1c and CV risk factors, but CV event or mortality risk was not positively impacted by the intervention [1]. No subgroups based on a single baseline characteristic were identified that showed a significant treatment effect on CV outcomes. It is possible, however, that a combination of patient characteristics affects the effect of an intervention.

This post hoc analysis of the Look AHEAD trial therefore aimed to explore whether the intensive lifestyle intervention may be beneficial in reducing CV events in individual patients.

To explore the possible presence of heterogeneity of treatment effect (HTE), treatment effect (prediction of MACE) was first modeled with a Cox proportional hazards model based on the Look AHEAD population. The choice of predictors was based on a novel lifetime risk model in patients with T2DM [5]. To model treatment effect directly, treatment-by-covariate interaction terms for all predictors were added to the model. The predictive value of the model was assessed based on calibration plots of predicted vs. observed 10-year risks of MACE and the c-statistic was used as a measure for discrimination.

The new prediction model was then used to estimate 10-year risk of MACE for each study participant, both as if they had been randomized to the intervention, as well as to the control group. The treatment effect was defined as the 10-year risk with diabetes support and education minus the 10-year risk with the intensive lifestyle intervention.

Then, HTE for intensive lifestyle intervention was assessed by dividing the study population into quartiles based on predicted treatment effect. Characteristics associated with a high vs. low predicted benefit for the risk of MACE were identified.

Main results

• The internal validation of the individual risk prediction model showed good agreement between the predicted and observed risk of MACE, with a c-statistic for discrimination of 0.73 (95%CI: 0.71-0.75).
• The median estimated baseline (untreated) 10-year risk for MACE was 15% (range: 0.3 – 96%). The median estimated absolute treatment effect of the lifestyle intervention on 10-year risk for MACE varied substantially, ranging from 10 year risk of -39% to +43% (median: -1.3%).
• Cox proportional hazard models adjusted for all prognostic factors included in the risk model showed an observed benefit of intervention in quartile 1 (HR: 0.64, 95%CI: 0.49-0.83) and no significant effect of the intervention in quartiles 2 and 3 (HR: 0.81, 95%CI: 0.58-1.14 and HR: 1.13, 95%CI: 0.80-1.60, respectively). In quartile 4, a harmful effect of intervention was seen (HR: 1.37, 95%CI: 1.09-1.37).
• Several patient characteristics varied significantly between quartiles of predicted treatment effect, including demographic aspects, medical history and medication use, and characteristics measured in physical examination and laboratory tests.
• When comparing quartiles 1 and 4 with the highest and lowest predicted benefit from intervention, respectively, overall, those in the first quartile had lower weight, lower systolic blood pressure, quite similar waist circumference, lower LDL-c and higher HbA1c than those in the fourth quartile.

Conclusion

References

Find this article online at Diabetes Care