Preprint: in 3067 patients who underwent cancer surgery, blood-based PhenoAge “biological age” predicted survival as well as a simple sum of the same blood measurements, and its excess over chronological age increased with tumor stage in three of four cancer types
Preprint: in 3067 patients who underwent cancer surgery, blood-based PhenoAge “biological age” predicted survival as well as a simple sum of the same blood measurements, and its excess over chronological age increased with tumor stage in three of four cancer types
On October 19, researchers at Memorial Sloan Kettering Cancer Center posted a study examining what PhenoAge “biological age” reflects on medRxiv, a server for manuscripts that have not yet undergone peer review. The cohort included patients with colon, kidney, lung, and breast cancer, with specific subtypes represented for the latter two. All had undergone tumor removal without receiving drug treatment before surgery.
Aging researcher Morgan Levine and colleagues introduced PhenoAge in 2018. It combines chronological age with nine blood measurements, weighted to predict mortality using data from the US National Health and Nutrition Examination Survey. A value of 55 in a 50-year-old indicates the same mortality risk as that of an average 55-year-old; the difference from chronological age is called PhenoAge acceleration. The preprint authors used eight measurements because C-reactive protein is rarely measured before surgery.
A study published in Nature Medicine in June used PhenoAge acceleration as a measure of aging and found it associated with the risk of cancer diagnosed before age 55. In someone with a tumor, a high PhenoAge could reflect both aging and disease: cancer causes inflammation and wasting that affect these same blood measurements, and the researchers who developed PhenoAge found in 2018 that it increased with the number of diseases a person had. In the preprint, patients with high PhenoAge acceleration had shorter survival in every group, although the association was not statistically significant in the lung cancer group, where 22 people died. The authors tested four predictions of the hypothesis that the measure reflects aging.
A measure of aging should be independent of tumor stage in patients who have undergone surgery. PhenoAge acceleration increased with T stage, which describes tumor size and extent. Correlations were 0,14 to 0,20 (the maximum is 1) in colon, kidney, and lung cancer. In breast cancer, the correlation was 0,06: 519 of 525 tumors were T1 or T2, leaving little variation in stage. Adjusting for T stage alone, a crude measure of disease burden, weakened the association between PhenoAge acceleration and survival by 10 to 17 % in the colon, kidney, and lung cancer groups.
The measure should improve predictions based on clinical data such as age, stage, and other diseases. Predictive performance is assessed using the C-index: the proportion of patient pairs for which the prediction correctly identifies who will die first (0,5 corresponds to chance and 1 to perfect prediction). Clinical data alone yielded values of 0,64 to 0,80; adding PhenoAge acceleration increased them by only 0,013 to 0,023.
The full measure should predict survival better than any individual blood measurement it contains. In three groups, however, predictive performance depended on one or two measurements, which differed by cancer type. In colon cancer, albumin alone (a measure of nutritional status) yielded a C-index of 0,69, compared with 0,68 for PhenoAge acceleration. In kidney cancer, lymphocyte percentage (a measure of immune status) yielded 0,69, compared with 0,71.
The weights derived from the general population should contribute prognostic information. The authors replaced each weight with its sign, assigning plus or minus one, and summed the eight blood measurements without including age. In every group, the C-index for this equally weighted sum was at least as high as that for PhenoAge acceleration. “The weights required for the formula to function as a biological clock add nothing to prediction in these groups,” the authors write.
The authors argue that this pattern is inconsistent with a general measure of aging: PhenoAge is sensitive to the systemic effects of cancer that appear in its component blood measurements. They allow that it may also capture an aging signal, but studying that signal in oncology requires measures whose signal can be distinguished from the effects of cancer and its treatment.