Calcium intake and bone mineral density as an example of non-linearity and threshold analysis.

UNLABELLED: Non-linearity is a likely phenomenon in bone metabolism, but is often ignored in pertinent epidemiological studies. Using NHANES III data on calcium intake and bone mineral density, the most important non-linear methods are introduced and discussed. The results should motivate researchers to consider non-linearity in this field more frequently. INTRODUCTION: Many relationships in bone metabolism and homeostasis are likely to follow non-linear patterns. Detailed dose-response analyses allowing for non-linear associations nonetheless remain scarce in this field. METHODS: A detailed analysis of NHANES III data on dietary calcium intake and bone mineral density was used to demonstrate the application and some of the challenges of the most important dose-response methods, including LOESS, categorical analysis, fractional polynomials, restricted cubic splines, and segmented regression. RESULTS: The spline estimate suggested increasing bone mineral density up to a calcium intake of about 1 g/day and a plateau thereafter. In segmented regression, the break-point marking the beginning of the plateau was placed at an intake of 0.58 (95 % confidence interval, 0.33 to 0.82) g/day. Sensitivity analyses suggested a less curved dose-response in women. CONCLUSIONS: Knowing about the possibilities and limitations of non-linear dose-response approaches should encourage researchers to consider these methods more frequently in studies on bone health and disease. The example analysis suggested bone mineral density to reach a plateau slightly below current calcium intake recommendations, with fairly pronounced differences of the dose-response shape by sex and menopausal status.