Nutrition biomarker analysis of COSMOS reveals novel approaches to improve the impact of randomized controlled trials in nutrition

Randomized Controlled Trials in nutrition (RCTN) are considered the gold standard for investigating the effects of diets and specific dietary compounds on human health.

However, RCTNs face unique challenges that are often overlooked. Different from randomized controlled trials testing drugs, RCTN are influenced by the background diet of participants and the challenge of assuring participants’ adherence to the dietary intervention under investigation.

In contrast, nutritional biomarkers are valuable tools in nutrition research, as they provide objective data on the intake of dietary compounds. As such, nutritional biomarkers can be used to assess the impact of background diet and adherence in RCTN.

In collaboration with the University of Reading and University of Cambridge, we previously developed and validated nutritional biomarkers to assess flavanol intake. Our partnership with Brigham and Women’s Hospital on the Cocoa Supplement and Multivitamin Outcomes Study (COSMOS) provided a unique opportunity to explore challenges in RCTNs by using flavanol biomarkers to objectively assess the impact of background diet and adherence on the outcomes of COSMOS.

What did we look into and why?

Background diet of participants and adherence to dietary interventions can significantly influence the outcomes of RCTNs. Participants in RCTNs continually consume food constituents and nutrients as part of their daily diets that are indistinguishable from those tested in the study intervention.

Very often this is not quantified by trial investigators and is not included in the analysis of outcomes. Additionally, assessing participant adherence to the dietary intervention of the trial traditionally relies on subjective, self-reported data that are prone to yield inaccurate results. Together, these two challenges can reduce the chance to detect changes in the outcomes measured in RCTNs.

As a subgroup of 6,532 participants from the COSMOS trial had provided biosamples for analysis, we had the opportunity to use flavanols’ biomarkers to objectively investigate the degree by which flavanol intake influence the outcomes of a large-scale RCTN such as COSMOS.

What did we do?

Using the flavanol biomarkers, we assessed flavanol intake in participants before and after they started to take the study intervention, i.e. the placebo or the 500mg of cocoa flavanols tested in COSMOS. By assessing both before and after the study intervention, we gained insights into participant’s previous intake of flavanols from background diet and whether they adhered to the study intervention. Cardiovascular (CVD) events experienced by participants were also recorded during participation in the study.

We then assessed the effect of cocoa flavanol intervention on the number of cardiovascular events in three different ways:

• Intention to treat (ITT) assessment: analysis of outcomes in participants taking the placebo vs intervention of 500mg of cocoa flavanols regardless of whether participants took the intervention or not. Here flavanol intake from background diet and adherence to cocoa flavanol intervention are not considered.

• Per protocol (PP) assessment: analysis of outcomes in participants taking the placebo vs cocoa flavanol intervention excluding those who reported they did not take the cocoa flavanol intervention. This analysis subjectively assesses adherence through self-reported diaries

• Biomarker-based assessment: analysis of outcomes in participants who had consumed more than 500mg of flavanols either from their diet or adherence to the cocoa flavanol intervention vs those who did not achieve 500mg flavanols based on an objective measure of intake using the biomarker. This classification of participants is regardless of which trail arm they were assigned to (placebo or intervention).

Both Intention to Treat and Per Protocol are typically used to analyse outcomes for RCTN.

What did we find?

We found that background diet and adherence have a substantial impact on the results of RCTNs:

• 1 in 5 participants already consumed flavanols as part of their background diet in amounts that are similar to those present in the cocoa flavanol intervention, and only 1 in 20 were not consuming any flavanols from their diet.

• Approximately 1 in 3 participants didn’t achieve the amount of flavanols set out for a full adherence to the cocoa flavanol intervention.

Comparatively, the group of participants who consumed more than 500mg of flavanols based on estimations with the flavanol biomarkers achieved a greater reduction in relative risk for total CVD events, CVD deaths, and major CVD events compared to the reductions assessed using the standard analyses implemented in RCTNs.

Table of results reflecting the table in the paper’s graphic abstract.

* Statistically significant

What does this mean?

The results obtained using flavanol biomarkers to objectively assess flavanol intake from background diet and adherence to the cocoa flavanol intervention confirm and strengthen the findings of COSMOS. Additionally, our findings indicate that flavanols could have an even more profound effect on reducing the risk of cardiovascular outcomes than the effect determined using the standard ITT or PP analyses.

For RCTNs, this means that background diet and adherence to study intervention can significantly impact trial outcomes if not properly accounted for. However, nutritional biomarkers can directly address the impact of these factors, providing a new approach to analyse RCTN outcomes.

Incorporating nutritional biomarkers in RCTNs can increase the capacity to detect meaningful changes in study outcomes following dietary interventions, and thus, increase the identification of nutrition strategies for maintaining and improving health.