Background: An evidence-based evaluation of peer-reviewed original research published during 1980 to 2004 and examining the relationship between hemoglobin and/or hematocrit values and all-cause mortality in dialysis patients was conducted to compare the studies' designs, analytic strategies, and results.
Methods: The search targeted MEDLINE and EMBASE and included publications referenced in the National Kidney Foundation-Kidney Disease Outcomes Quality Initiative Anemia Guidelines. Both randomized clinical trials (RCTs) and observational studies were considered.
Results: Of 7 RCTs and 20 observational studies identified, 5 trials and 13 studies were included in evidence tables. The trials were underpowered to study mortality and enrolled different patient populations, limiting their generalizability. Although none reached statistical significance, trials focusing on a general dialysis population tended to show either no effect or a benefit of greater hemoglobin level target, whereas trials enrolling cardiac patients suggested increased mortality associated with greater hemoglobin levels. The observational studies were heterogeneous in design, used varying exposure categorizations, and controlled for different covariates, but generally were supportive of increased mortality associated with a hemoglobin level less than the reference range. Evidence of benefits or risks of hemoglobin levels greater than the reference was variable.
Conclusion: RCT-based evidence relating hemoglobin and/or hematocrit values to mortality in dialysis patients is limited. The relationship may be modified by the presence of preexisting conditions (cardiac disease). The published literature is insufficient for generalization of risks or benefits of a hemoglobin level greater than 11 to 12 g/dL (>110 to 120 g/L). There is a need for better designed RCTs focusing on mortality as a primary outcome and enrolling patients without cardiac disease. Observational studies should adequately control for relevant confounders (eg, baseline comorbidities) and assess effect modification in the analysis.