J

A depiction of a non-linear relationship between a quantitative, ordinal or ordered categorical risk factor and risk or probability of disease (or death), whereby the lowest levels of exposure are associated with an increased risk as compared to low to moderate levels of exposure to the risk factor in question. As extent of exposure increases from this point, a sharp or more gradual increase in risk occurs, giving the characteristic J-shape (in other words, a J-shaped curve has one turn point). The opposite pattern characterizes a reverse J-shaped curve: highest risk for low levels of a given factor (likely, a preventive factor), lowest risk at moderate to high levels of the factor, with the highest levels of the factor again experiencing an increase in risk (Figure J.1 b - SEMPOS et al, 2013). A J-shaped curve has been described for the relationship between ethanol intake, blood pressure measurements and serum cholesterol levels and risk of cardiovascular disease (CVD). Under this pattern of relationship, alcohol abstainers seem to have greater risk of CVD than light to moderate drinkers, and very low blood pressure and cholesterol levels likewise do not appear to 176confer protection. This is in contrast, for example, with exposure to cigarette smoke and risk of C VD and lung cancer, where non-smokers experience the lowest risk of disease and mortality rates. Research studies have alternatively found J-shaped relationships to have, rather, a U-shaped pattern. In either case, interpretation of these effects requires careful consideration of possible harms vs. potential benefits. Figure J.1 a (MIDLOV et al., 2016) illustrates the J-shaped relationship between self-reported alcohol intake (in g/day) and all-cause mortality (expressed as hazard ratios by alcohol consumption), in a follow-up study involving 6353 women aged 50-59 years, in the Lund area in Sweden (participation rate was 64.2%, 6916 women, of whom 563 were excluded due to incomplete data). Average duration of follow-up was 17 years, and 579 deaths were registered by the end of the study period. Hazard ratios are adjusted for sociodemographic and lifestyle factors, diabetes and previous ischaemic heart disease. See also di CASTELNUOVO et al. (2006); dose- response relationship. Cf. exponential curve.