Are U‐shaped relationships between risk factors and outcomes artifactual?

Abstract

Background: The objective of this study was to evaluate whether the observed nadir in a U-or J-shaped relationship between a particular risk factor and a future health outcome is a function of the distribution of the risk factor in the sample being analyzed.Methods: Data from the ORIGIN trial were used to assess the relationship between three risk factors (weight, systolic blood pressure, and serum insulin) and the hazard of a major cardiovascular event comprising a nonfatal myocardial infarction, nonfatal stroke, or cardiovascular death. Three spline curves were generated for each risk factor. The first was based on all available data, the second for a subgroup with a higher mean risk factor level, and the third for a subgroup with a lower mean risk factor level. Nadir levels of the risk factor (i.e., risk factor levels predicting the lowest hazard) were then identified for each spline curve.Results: When compared to the nadir values based on all available data, nadir values for all three risk factors were higher for the subgroups with higher mean levels and lower for those with lower mean levels. Conclusions: The distribution of a risk factor in the population is an important determinant of its nadir value. Populations with high or low values may have high and low nadirs, respectively. Identification of a nadir for a modifiable risk factor from epidemiologic relationships may therefore arise from this distribution bias and is therefore unrelated to therapeutic targets.

Stress hyperglycemia ratio is associated with systemic inflammation and clinical outcomes in diabetic inpatients with pneumonia on admission

“…It was indicated in a previous study that the distribution of a risk factor in the study population greatly influenced its nadir value in a U‐ or J‐shaped relationship between a particular risk factor and a future health outcome, as well as the exact hazard ratio or OR 31 . Consequently, considering the sample size of this study, we did not point out the nadir value or make further investigation on the exact range.…”

“…It was indicated in a previous study that the distribution of a risk factor in the study population greatly influenced its nadir value in a U-or J-shaped relationship between a particular risk factor and a future health outcome, as well as the exact hazard ratio or OR. 31 Consequently, considering the sample size of this study, we did not point out the nadir value or make further investigation on the exact range. Further large-scale studies are needed to find this nadir value and favorable range of SHR and make them better serve in the prediction of adverse clinical outcomes in diabetic inpatients with pneumonia.…”