- Email: [email protected]
Target haemoglobin concentrations in chronic kidney disease
Correspondence
4
5
Regidor DL, Kopple JD, Kovesdy CP, et al. Associations between changes in hemoglobin and administered erythropoiesis-stimulating agent and survival in hemodialysis patients. J Am Soc Nephrol 2006; 17: 1181–91. Strippoli GF, Tognoni G, Navaneethan SD, Nicolucci A, Craig JC. Haemoglobin targets: we were wrong, time to move on. Lancet 2007; 369: 346–50.
On reading the report of the metaanalysis by Arintaya Phrommintikul and colleagues1 on haemoglobin targets in patients with chronic kidney disease (CKD), we wondered why the authors decided to pool studies in CKD stage 3–4 patients with studies done in haemodialysis patients, since dialysis is known to increase cardiovascular events and mortality by threefold to fivefold. Given the heterogeneity and limited number of the studies included, standard methods for analysis are not appropriate and mixed-effect models should be used. Although a sensitivity analysis was done, which yielded nonsignificant results, these findings are not reflected in the conclusions. In the accompanying Comment by Giovanni Strippoli and colleagues,2 the statement that in the nine trials in the meta-analysis “all outcomes… favoured lower, rather than higher, haemoglobin target concentrations” is incorrect. In the CREATE trial,3 there was no significant difference between lower and higher haemoglobin targets for cardiovascular outcomes and mortality, especially when censoring patients at the start of dialysis. The CHOIR study,4 although it censored patients on dialysis and was stopped prematurely, only found a borderline difference in outcomes favouring the lower haemoglobin target. Of note, in CREATE, several quality-of-life variables were significantly better in the high than in the low haemoglobin target group, in line with previous reports in CKD patients.5 Therefore, we are not sure that the issue of the optimum haemoglobin target concentration is really settled. TB Drüeke receives consulting fees and lecture fees from F Hoffman-La Roche and consulting fees, lecture fees, and grant support from Amgen.
1516
Tilman B Drüeke, on behalf of the CREATE Steering Committee [email protected] INSERM Unite 845, Hôpital Necker, Paris, France 1
2
3
4
5
Phrommintikul A, Haas SJ, Elsik M, Krum H. Mortality and target haemoglobin concentrations in anaemic patients with chronic kidney disease treated with erythropoietin: a meta-analysis. Lancet 2007; 369: 381–88. Strippoli GFM, Tognoni G, Navaneethan SD, Nicolucci A, Craig JC. Haemoglobin targets: we were wrong, time to move on. Lancet 2007; 369: 346–49. Drüeke TB, Locatelli F, Clyne N, et al. Normalization of hemoglobin in patients with chronic kidney disease. N Engl J Med 2006; 355: 2071–84. Singh AK, Szczech L, Tang KL, et al. Correction of anemia with epoetin alfa in chronic kidney disease. N Engl J Med 2006; 355: 2085–98. Levin A. Understanding recent haemoglobin trials in CKD: methods and lesson learned from CREATE and CHOIR. Nephrol Dial Transplant 2007; 22: 309–12.
Authors’ reply Julie Ingelfinger and Giuseppe Remuzzi (as well as Deborah Regidor and Kamyar Kalantar-Zadeh and Tilman Drüeke and colleagues) question whether patients with chronic kidney disease (CKD), both before dialysis and on dialysis, should be considered together in a single meta-analysis. We consider, as do relevant guideline bodies,1 that these states represent a continuum within the same condition and therefore should be considered together within an overall assessment of the effect of treatment. Ingelfinger and Remuzzi were actually incorrect in stating that only two studies in our metaanalysis merely involved predialysis patients—there were six in which predialysis was at least a component of the study cohort. Regidor and Kalantar-Zadeh also express concern about our dichotomous grouping of haemoglobin targets (high vs low). These merely reflect the grouping of haemoglobin targets within the individual studies. Multiple studies often show heterogeneity across populations and outcomes. This is one of the potential drawbacks of meta-analysis, as specifically alluded to in our discussion of
the study’s limitations. Nevertheless, populations were actually quite similar across haemoglobin groups in the studies analysed. Similar odds ratios for major events including death were seen in our analyses, whether or not patients were undergoing dialysis. Drüeke and colleagues suggest that fixed-effects analysis is not appropriate when significant heterogeneity is seen; we concur, and had applied random-effects modelling when such heterogeneity was seen in our analyses. This is not an adjustment for heterogeneity, contrary to the comments made by Regidor and Kalantar-Zadeh. We did not suggest that research should not continue in this area. Indeed, we noted that TREAT might partly answer the question of the optimum haemoglobin concentration in such patients. TREAT will not, however, be able to address the main issue raised by our meta-analysis—ie, an increase in all-cause mortality at higher haemoglobin concentrations— because it is underpowered to make such an assessment. We agree with Regidor and Kalantar-Zadeh that both adequacy of haemodialysis and use of intravenous iron therapy might be confounding factors in our analysis. Of the studies that reported these data,2,3 both haemoglobin target groups had a mean dialysis dose of more than 1·2 sKt/V, which represents adequate dialysis according to the Kidney Disease Outcomes Quality Initiative’s clinical practice guideline.4 There is much controversy about the effect of intravenous iron on clinical outcomes in patients with CKD and other patients. Although intravenous iron treatment is strongly recommended in haemodialysis patients to improve haemoglobin concentrations and reduce erythropoietin requirements, there have thus far been no iron intervention trials sufficiently powered to assess safety and longterm clinical outcomes.1 Furthermore, intravenous iron increases oxidative www.thelancet.com Vol 369 May 5, 2007
4
5
Regidor DL, Kopple JD, Kovesdy CP, et al. Associations between changes in hemoglobin and administered erythropoiesis-stimulating agent and survival in hemodialysis patients. J Am Soc Nephrol 2006; 17: 1181–91. Strippoli GF, Tognoni G, Navaneethan SD, Nicolucci A, Craig JC. Haemoglobin targets: we were wrong, time to move on. Lancet 2007; 369: 346–50.
On reading the report of the metaanalysis by Arintaya Phrommintikul and colleagues1 on haemoglobin targets in patients with chronic kidney disease (CKD), we wondered why the authors decided to pool studies in CKD stage 3–4 patients with studies done in haemodialysis patients, since dialysis is known to increase cardiovascular events and mortality by threefold to fivefold. Given the heterogeneity and limited number of the studies included, standard methods for analysis are not appropriate and mixed-effect models should be used. Although a sensitivity analysis was done, which yielded nonsignificant results, these findings are not reflected in the conclusions. In the accompanying Comment by Giovanni Strippoli and colleagues,2 the statement that in the nine trials in the meta-analysis “all outcomes… favoured lower, rather than higher, haemoglobin target concentrations” is incorrect. In the CREATE trial,3 there was no significant difference between lower and higher haemoglobin targets for cardiovascular outcomes and mortality, especially when censoring patients at the start of dialysis. The CHOIR study,4 although it censored patients on dialysis and was stopped prematurely, only found a borderline difference in outcomes favouring the lower haemoglobin target. Of note, in CREATE, several quality-of-life variables were significantly better in the high than in the low haemoglobin target group, in line with previous reports in CKD patients.5 Therefore, we are not sure that the issue of the optimum haemoglobin target concentration is really settled. TB Drüeke receives consulting fees and lecture fees from F Hoffman-La Roche and consulting fees, lecture fees, and grant support from Amgen.
1516
Tilman B Drüeke, on behalf of the CREATE Steering Committee [email protected] INSERM Unite 845, Hôpital Necker, Paris, France 1
2
3
4
5
Phrommintikul A, Haas SJ, Elsik M, Krum H. Mortality and target haemoglobin concentrations in anaemic patients with chronic kidney disease treated with erythropoietin: a meta-analysis. Lancet 2007; 369: 381–88. Strippoli GFM, Tognoni G, Navaneethan SD, Nicolucci A, Craig JC. Haemoglobin targets: we were wrong, time to move on. Lancet 2007; 369: 346–49. Drüeke TB, Locatelli F, Clyne N, et al. Normalization of hemoglobin in patients with chronic kidney disease. N Engl J Med 2006; 355: 2071–84. Singh AK, Szczech L, Tang KL, et al. Correction of anemia with epoetin alfa in chronic kidney disease. N Engl J Med 2006; 355: 2085–98. Levin A. Understanding recent haemoglobin trials in CKD: methods and lesson learned from CREATE and CHOIR. Nephrol Dial Transplant 2007; 22: 309–12.
Authors’ reply Julie Ingelfinger and Giuseppe Remuzzi (as well as Deborah Regidor and Kamyar Kalantar-Zadeh and Tilman Drüeke and colleagues) question whether patients with chronic kidney disease (CKD), both before dialysis and on dialysis, should be considered together in a single meta-analysis. We consider, as do relevant guideline bodies,1 that these states represent a continuum within the same condition and therefore should be considered together within an overall assessment of the effect of treatment. Ingelfinger and Remuzzi were actually incorrect in stating that only two studies in our metaanalysis merely involved predialysis patients—there were six in which predialysis was at least a component of the study cohort. Regidor and Kalantar-Zadeh also express concern about our dichotomous grouping of haemoglobin targets (high vs low). These merely reflect the grouping of haemoglobin targets within the individual studies. Multiple studies often show heterogeneity across populations and outcomes. This is one of the potential drawbacks of meta-analysis, as specifically alluded to in our discussion of
the study’s limitations. Nevertheless, populations were actually quite similar across haemoglobin groups in the studies analysed. Similar odds ratios for major events including death were seen in our analyses, whether or not patients were undergoing dialysis. Drüeke and colleagues suggest that fixed-effects analysis is not appropriate when significant heterogeneity is seen; we concur, and had applied random-effects modelling when such heterogeneity was seen in our analyses. This is not an adjustment for heterogeneity, contrary to the comments made by Regidor and Kalantar-Zadeh. We did not suggest that research should not continue in this area. Indeed, we noted that TREAT might partly answer the question of the optimum haemoglobin concentration in such patients. TREAT will not, however, be able to address the main issue raised by our meta-analysis—ie, an increase in all-cause mortality at higher haemoglobin concentrations— because it is underpowered to make such an assessment. We agree with Regidor and Kalantar-Zadeh that both adequacy of haemodialysis and use of intravenous iron therapy might be confounding factors in our analysis. Of the studies that reported these data,2,3 both haemoglobin target groups had a mean dialysis dose of more than 1·2 sKt/V, which represents adequate dialysis according to the Kidney Disease Outcomes Quality Initiative’s clinical practice guideline.4 There is much controversy about the effect of intravenous iron on clinical outcomes in patients with CKD and other patients. Although intravenous iron treatment is strongly recommended in haemodialysis patients to improve haemoglobin concentrations and reduce erythropoietin requirements, there have thus far been no iron intervention trials sufficiently powered to assess safety and longterm clinical outcomes.1 Furthermore, intravenous iron increases oxidative www.thelancet.com Vol 369 May 5, 2007