Risk factors for higher mortality at the highest levels of spKt/V in haemodialysis patients.

BACKGROUND The survival of patients on haemodialysis improves as the delivered doses of dialysis attain a Kt/V of 1.2 or more. However, a consistent yet paradoxical finding in the Kt/V survival relationship is that the mortality tends to increase at the higher ends of Kt/V. METHOD To determine the relationship of Kt/V with survival at a time when increasing doses of dialysis are being delivered and to examine the effect of body mass and nutritional markers including serum pre-albumin on the paradoxical relationship, we analysed relative mortality risk (RR) as a function of single-pool Kt/V (spKt/V) in the Cox proportional hazard model. We used body mass index (BMI), dry body weight and nutritional parameters including serum pre-albumin obtained in 1151 patients on chronic haemodialysis as covariates. RESULTS The mean spKt/V for February and March of 1997 was 1.46+/-0.28 (+/-SD). A spKt/V of >1.2 was achieved in 82.5% of patients and 20% of patients received a spKt/V of >1.68. Using spKt/V in deciles and assigning the third decile (spKt/V 1.2-1.3) as the reference group with an RR of 1, the extreme first and the tenth deciles displayed a higher RR imparting a U-shaped configuration to the curve. In this unadjusted analysis, there was no dependency between delivered dose of spKt/V and RR values. spKt/V values were re-analysed in quintiles. The U-shaped relationship persisted between spKt/V and RR, and an unadjusted analysis again exhibited no clear dependency between spKt/V and RR. The patients in the highest fifth spKt/V quintile, who received the highest dose of dialysis and had the paradoxical increase in RR, had the lowest body weight, BMI, serum pre-albumin and creatinine. Adjustments for case-mix characteristics (age, gender, race and diabetes) in the Cox multivariate model did not reduce the paradoxical increase in RR. However, introduction of BMI or dry body weight along with serum creatinine and pre-albumin to the above case-mix covariates for the first time produced a dose-dependent inverse relationship between the first four quintiles of the spKt/V and their respective RR. With the above variables adjusted, the fifth quintile RR of 1.6 (0.9, 3.1) was reduced to 0.9 (0.4, 2.0), but was not corrected to the lowest RR of 0.6 (0.2, 1.2) noted in the fifth spKt/V quintile. This difference between the adjusted RR of fifth and fourth quintile was not statistically significant, but persisted after adjusting for any clustering variation. CONCLUSIONS Our analysis, which is the first to include serum pre-albumin in the Kt/V survival analysis, demonstrates a steeper rise in the unadjusted RR at the highest end of spKt/V levels than reported previously and suggests that patients with lower weight and nutritional parameters may mostly account for the spKt/V and RR paradox. As we find a worsening in the spKt/V-RR paradox at a time when higher doses of dialysis are being delivered, we speculate that factors other than underweight and malnutrition such as 'toxicity' of rapid dialysis, especially in sick and underweight patients, may contribute to the paradox. If future studies were to verify this possibility, sick and underweight patients could benefit from less vigorous but frequent sessions of haemodialysis.