Approved Research

Use of protein biomarkers to estimate the Healthy Lifestyle Index and morbidity and mortality risk (Submitted on behalf of Health Outlook Corp.)

Lay summary

Both genetic and lifestyle factors affect an individual's risk of developing life-threatening illnesses such as cardiovascular disease and cancer. To study the relationship between lifestyle and disease, scientists have developed a score called the Healthy Lifestyle Index (HLI), based on diet quality, alcohol consumption, smoking, physical activity, and body mass index or waist circumference. Consistent with the importance of lifestyle being a determinant of disease risk, a high HLI score has been associated with reduced risk of various cancers and cardiovascular disease. However, a significant limitation of the HLI which has prevented it from being used outside the research setting is the time taken to collect the requisite information. Assessing factors such as diet quality and physical activity requires significant time and effort, making this impractical outside of a research setting. A test based on age, sex and blood biomarkers that mirrored the HLI score would be potentially transformative.

There is growing evidence that signatures based on blood levels of proteins (proteomics) reflect underlying disease and mortality risk. Here we posit that it will be possible to develop a practical test that includes multiple blood proteins that correlates with the HLI score. Indeed, we already know of protein biomarkers in blood that reflect components of the HLI. For example, increased waist size is associated with elevated levels of C-reactive protein, a marker of inflammation. Currently, a company called Olink is carrying out an analysis of approximately 3,000 proteins in blood from about 57,000 UK Biobank participants. Once these data become available, we propose to undertake analyses to test our hypothesis. Specifically, we will build and verify a statistical model that includes basic demographic variables (age, sex) as well as blood protein markers to predict the HLI score. We will then use the scores derived from the model to examine their association with risk of cardiovascular disease, selected cancers, and death. Further, we will compare the performance of this model for predicting mortality risk with that of a more comprehensive model of mortality risk that includes classical mortality risk factors (e.g., age, sex, body mass index, cigarette smoking history, etc.) and blood biomarkers.

Potentially, the project described here might lead to the development of a pragmatic HLI test that could be used to personalize healthcare recommendations or to monitor the efficacy of interventions designed to improve the health of a work force or population.

Scope extension:

1. To develop/test a pragmatic multivariable model comprised of age, sex, and blood biomarkers (including proteins) for estimation of the healthy lifestyle index (HLI);
2. Examine whether scores derived from the multivariable model developed in aim 1 are associated with risk of disease (CVD, breast cancer)/hospitalization/all-cause mortality; 
3. Develop/test predictors of all-cause mortality based on classical mortality risk factors (e.g.,age, sex, body mass index) and blood biomarkers (including proteins);
4. Compare the performance with respect to prediction of all-cause mortality of the best model identified in aim 3 with that of the HLI estimation model.
   In our initial analysis of the proteomic data, we have identified both clinical factors and a large number of proteins that are associated with all-cause mortality. In particular, high levels of GDF15 and WFDC2 proteins were associated with elevated mortality risk. We now propose to extend this line of investigation to disease-specific mortality.  For example, it will be of interest to determine which blood-derived proteins are associated with risk of death from cardiovascular disease, cancer, respiratory diseases, etc.  To this end, we will use the same study population that we proposed to use for our initial submission (i.e., the more than 50,000 subjects who had proteomic analyses performed using baseline blood samples).  In the nearly 4,000 subjects who died prior to March 2020, we will categorize those who died by disease-specific causes of death. Using a similar analytic approach to that described in our original proposal, levels of proteins at baseline will then be compared in subjects who died from a specific cause versus those who were alive at the end of the follow-up period.  These analyses should permit us to determine whether or not there are specific sentinel proteins that reflect elevated risk for specific causes of death.