Sex-specific association of time-varying haemoglobin values with mortality in incident dialysis patients.

BACKGROUND Previous studies in dialysis patients showed an association between haemoglobin levels and all-cause mortality, however, without addressing sex differences. METHODS We followed up 235 incident dialysis patients of the region of Vorarlberg in a prospective cohort study applying a time-dependent Cox regression analysis using all the measured laboratory values for up to more than 7 years. In total, 12 242 haemoglobin measurements with a median of 47 (range 3-270) per patient were available to evaluate the impact of haemoglobin levels and their variability on all-cause mortality in a sex-stratified analysis. Non-linear P-splines were used to allow a flexible modelling of the association with mortality. RESULTS We observed an inverse relationship between the increasing haemoglobin values and the decreasing risk of mortality. The linear component of the non-linear spline was highly significant for both men (P = 0.00005) and women (P = 0.0000000052). The non-linear component was also significant but less pronounced than the linear component. The inverse relationship was clear to see haemoglobin levels of up to 12-13 g/L in women, which reached a plateau for the higher values of haemoglobin. For men, an inverse trend was observed but clearly attenuated when compared to women. After adjustment for additional parameters of inflammation and malnutrition as well as diabetes, the linear component was more significant in women (P = 0.0018) than in men (P = 0.023). CONCLUSIONS This study applied for the first time a time-dependent Cox regression analysis over a long-term observation period of several years using all available measurements. Besides the methodological advantages, our data indicate a sex-specific linear as well as non-linear effect of haemoglobin levels on all-cause mortality, which was markedly more pronounced in women.