Commonly used clinical chemistry tests as mortality predictors: Results from two large cohort studies

Background The normal ranges for clinical chemistry tests are usually defined by cut-offs given by the distribution in healthy individuals. This approach does however not indicate if individuals outside the normal range are more prone to disease. Methods We studied the associations and risk prediction of 11 plasma and serum biomarkers with allcause mortality in two population-based cohorts: A Swedish cohort (X69) initiated in 1969, and the UK Biobank (UKB) initiated in 2006-2010, with up to 48- and 9-years follow-up, respectively. Results In X69 and in UKB, 18,529 and 425,264 individuals were investigated, respectively. During the follow-up time, 14,475 deaths occurred in X69 and 17,116 in UKB. All evaluated tests were associated with mortality in X69 (P<0.0001, except bilirubin P<0.005). For calcium, blood urea nitrogen, bilirubin, hematocrit, uric acid, and iron, U-shaped associations were seen (P<0.0001). For leukocyte count, gamma-glutamyl transferase, alkaline phosphatases and lactate dehydrogenase, linear positive associations were seen, while for albumin the association was negative. Similar associations were seen in UKB. Addition of all biomarkers to a model with classical risk factors improved mortality prediction (delta C-statistics: +0.009 in X69 and +0.023 in UKB, P<0.00001 in both cohorts). Conclusions Commonly used clinical chemistry tests were associated with all-cause mortality both in the medium- and long-term perspective, and improved mortality prediction beyond classical risk factors. Since both linear and U-shaped relationships were found, we propose to define the normal range of a clinical chemistry test based on its association with mortality, rather than from the distribution.