Skip to main content
Premium Trial:

Request an Annual Quote

Decode Study Identifies Proteomic Predictors of All-Cause Mortality

NEW YORK – A team led by researchers at Amgen subsidiary Decode Genetics have used proteomic measurements to develop predictors for short- and long-term risk of all-cause mortality.

Detailed in a study published this week in Nature Communications Biology, the predictors were able to identify a high-risk group of subjects between 60 and 80 years old, 88 percent of whom died within 10 years, as well as a low-risk group in which 1 percent died within 10 years.

The researchers also identified protein profiles specific to various causes of death, finding in particular that deaths linked to the nervous system differed in the proteins involved from most other causes of death.

To build their predictors, the Decode team measured the levels of 4,684 proteins in 22,913 individuals between 18 and 101 years of age, who they followed for a mean period of 13.7 years. During that follow-up period 7,061 of the participants died. They developed their models using 70 percent of the subject data and then tested those models using the remaining 30 percent.

The predictors consisted of subject age, sex, and between 81 and 219 protein measurements, which outperformed a predictor composed of traditional risk factors both in terms of short-term and long-term risk prediction and for the full study cohort as well as for a subset including only participants over 60 years of age.

The most powerful single protein in terms of risk of death prediction was GDF15, a previously identified marker for mortality risk. A model consisting of age, sex, and GDF15 alone outperformed the model based on traditional risk factors, which no other single protein was able to do. The researchers found, however, that a combination of proteins could improve upon the traditional risk factor model in the absence of GDF15.

Using the protein models, the researchers identified a group comprising 5 percent of the 60- to 80-year-old cohort (124 subjects total) with an 88 percent chance of dying in 10 years and a 67 percent chance of dying in five years. Used to similarly identify a high-risk group, the traditional risk-factor model identified subjects with a 65 percent probability of dying in 10 years and a 40 percent probability of dying in five years.

The protein models identified another 5 percent of the 60- to 80-year-old cohort (125 subjects total) at the lowest risk of death, 1 percent of whom died within 10 years. Low-risk subjects identified by the traditional risk-factor model had a 5 percent chance of dying within 10 years.

The researchers also looked at an independent dataset at how their measurements matched with phenotypes — including performance on an exercise test and lean body mass — considered to be measures of health and frailty, finding that their predictors correlated well with these phenotypic measures.

The study "shows the power of protein levels in plasma as predictors of death," the authors wrote, adding that this finding suggests the potential of plasma proteomics to identify "causal relationships, and useful biomarkers for early detection of different health problems and the possibility of intervention."