Articles. Funding None.

Transcription

1 Effect of calcium-based versus non-calcium-based phosphate binders on mortality in patients with chronic kidney disease: an updated systematic review and meta-analysis Sophie A Jamal, Ben Vandermeer, Paolo Raggi, David C Mendelssohn, Trish Chatterley, Marlene Dorgan, Charmaine E Lok, David Fitchett, Ross T Tsuyuki Lancet ; 82: Published Online July 9, S4-676()6897- See Comment page Division of Endocrinology and Metabolism, Women s College Hospital (S A Jamal MD), Women s College Research Institute (S A Jamal), Division of Nephrology (C E Lok MD), and Division of Cardiology, St Michael s Hospital (D Fitchett MD), University of Toronto, Toronto, ON, Canada (Prof D C Mendelssohn MD); Alberta Research Centre for Health Evidence, Department of Pediatrics (B Vandermeer MSc), Mazankowski Alberta Heart Institute (Prof P Raggi MD), John W Scott Health Sciences Library, Walter C Mackenzie Health Sciences Centre (T Chatterley MLIS, M Dorgan MLS), and Division of Cardiology (Prof R T Tsuyuki PharmD), University of Alberta, Edmonton, AB, Canada; and Department of Nephrology, Humber River Regional Hospital, Toronto, ON, Canada (D C Mendelssohn) Correspondence to: Dr Sophie A Jamal, Women s College Hospital Research Institute, Toronto, ON, M5G N8, Canada sophie.jamal@utoronto.ca Summary Background Phosphate binders (calcium-based and calcium-free) are recommended to lower serum phosphate and prevent hyperphosphataemia in patients with chronic kidney disease, but their effects on mortality and cardiovascular outcomes are unknown. We aimed to update our meta-analysis on the effect of calcium-based versus non-calciumbased phosphate binders on mortality in patients with chronic kidney disease. Methods We did a systematic review of articles published in any language after Aug,, up until Oct, 2, by searching Medline, Embase, International Pharmaceutical Abstracts, Cochrane Central Register of Controlled Trials, and Cumulative Index to Nursing and Allied Health Literature. We included all randomised and non-randomised trials that compared outcomes between patients with chronic kidney disease taking calcium-based phosphate binders with those taking non-calcium-based binders. Eligible studies, determined by consensus with predefined criteria, were reviewed, and data were extracted onto a standard form. We combined data from randomised trials to assess the primary outcome of all-cause mortality using the DerSimonian and Laird random effects model. Findings Our search identified 847 reports, of which eight new studies (five randomised trials) met our inclusion criteria and were added to the ten (nine randomised trials) included in our previous meta-analysis. Analysis of the randomised trials (46 patients) that reported an outcome of mortality showed that patients assigned to noncalcium-based binders had a % reduction in all-cause mortality compared with those assigned to calcium-based phosphate binders (risk ratio 78, 95% CI 6 98). Interpretation Non-calcium-based phosphate binders are associated with a decreased risk of all-cause mortality compared with calcium-based phosphate binders in patients with chronic kidney disease. Further studies are needed to identify causes of mortality and to assess whether mortality differs by type of non-calcium-based phosphate binder. Funding None. Introduction Chronic kidney disease is a worldwide public health problem with a yearly incidence as high as 2 cases per million in many countries. Cardiovascular disease is the leading cause of death in men and women with chronic kidney disease, and accelerated, progressive vascular calcification might play a part in the development of this disorder. 2 The cause of vascular calcification is not known, but probably involves the conversion of vascular smooth muscle cells into bone forming, osteoblast-like, cells. Factors that promote this change in cell type include high concentrations of serum calcium and phosphorus. In fact, results of observational studies in patients with chronic kidney disease show an association between cardiovascular mortality and increases in serum calcium, phosphate, and calcium phosphate product. 4 Hyperphosphataemia is an important and almost inevitable clinical result of advanced chronic kidney disease, and controlling serum phosphate both by dietary restriction and use of an intestinal phosphate binder (calcium-based or non-calcium-based) is a key focus for clinicians. The possibility that non-calcium-based phos phate binders might by attenuating positive calcium balance delay the progression of vascular calcification and cardiovascular and overall mortality needs to be investigated. Results from individual clinical trials that have examined the relation between calcium-based phosphate binders, vascular calcification, and mortality have been inconclusive probably because of the small numbers of study participants and outcome events. 5 2 This limitation has left clinicians uncertain as to which type of phosphate binder should be prescribed in patients with chronic kidney disease. To address this uncertainty, in 29, we did a meta-analysis on the effects of calcium-based versus non-calcium-based phosphate binders on mortality in patients with chronic kidney disease. 24 On the basis of data from eight randomised controlled trials and 7 participants, we noted a trend towards a decrease in all-cause mortality in patients taking Vol 82 October 2,

2 non-calcium-based phosphate binders versus those taking calcium-based phosphate binders (risk ratio [RR] 68, 95% CI 4 ). We recognised that our findings were limited by low power, and as such could not exclude a potentially important beneficial effect. Thus, we formed a collaborative review group to undertake an updated meta-analysis. We aimed to update our meta-analysis on the effect of calcium-based versus non-calcium-based phosphate binders on mortality in patients with chronic kidney disease. Trials that have been published since our original meta-analysis are inconclusive as to the relation between type of phosphate binder and mortality. As such, we aimed to provide the best available evidence as to which type of phosphate binder clinicians should prescribe to patients with chronic kidney disease. Methods Search strategy and selection criteria We did a systematic review in accordance with the PRISMA (Preferred Reporting Items for Systematic reviews and meta-analyses) guidelines. 25 The search strategy was developed and undertaken by two experienced medical librarians (TC and MD), who searched for clinical trials published after Aug, (when our previous systematic review ended), up until Oct, 2 (the date of the last search). They searched the electronic databases Ovid Medline (articles published between 946 and October, 2, including those in the process of being added to the database and other citations from journals listed in PubMed but not included in Medline), Ovid Embase (974 to October, 2), Ovid International Pharmaceutical Abstracts (97 to September, 2), Wiley Cochrane Central Register of Controlled Trials (inception to October, 2), and EBSCO Cumulative Index to Nursing and Allied Health Literature (inception to October, 2). Keywords included phosphate binders, calcium, kidney disease, dialysis, cardiovascular events, and mortality. Additionally, one investigator (SJ) manually searched through the cited references of published reviews of phosphate binders in chronic kidney disease. We attempted contact with authors to clarify published data if needed. The complete search strategy is available in the appendix. We included all randomised and non-randomised trials published in any language that compared outcomes between patients (men or women irrespective of menopausal status) with chronic kidney disease (irrespective of stage of chronic kidney disease or type of dialysis) taking calcium-based phosphate binders (calcium carbonate or calcium acetate) and those taking non-calcium-based binders (sevelamer hydrochloride, sevelamer carbonate, or lanthanum). We excluded quasiexperiments, abstracts, and unpublished studies. The primary outcome was all-cause mortality. The secondary outcomes were cardiovascular events (fatal and non-fatal myocardial infarction, fatal and non-fatal stroke, and sudden death), vascular calcification, fractures, and vascular compliance. Data extraction We selected studies and extracted the data according to a standard Cochrane protocol. 25,26 All the investigators independently reviewed the abstracts and identified potential manuscripts for retrieval. We established study eligibility by consensus, using previously defined inclusion criteria. Two investigators (SJ and RT) independently reviewed eligible studies for study characteristics and clinical relevance and, if appropriate, extracted study data. We used consensus and a third reviewer (PR), if necessary, to resolve disagreements. The following information was extracted onto standardised data collection forms: number of patients included (pooled and by treatment assignment), number of patients excluded, number of patients observed, number of patients lost to follow-up and the reasons for loss to follow-up, population characteristics (age, weight, sex, and, for women, menopausal status), stage and duration of chronic kidney disease, presence of diabetes or hypertension, smoking status, use of cardiac medication or bisphosphonates, and laboratory results at randomisation. We also extracted data on trial characteristics (inclusion and exclusion criteria), type of study (for randomised controlled trials we recorded independent randomisation centre, blindness, random allocation, adequate allocation concealment, intention to treat, and withdrawal or dropout rate), and trial intervention. Our data abstraction form is available upon request. Risk of bias assessment We used the Cochrane risk of bias method to appraise study quality; two reviewers (SJ and RT) independently rated each study on the six domains (low, unclear, or high bias for sequence generation; allocation concealment; blinding of participants, personnel and outcome assessors; incomplete outcome data; selective outcome reporting; and other sources of bias) and then compared assessments. Any discrepancies were resolved by 847 potentially relevant articles identified and screened for retrieval 48 articles retrieved for more detailed assessment 8 articles included in the updated meta-analysis Figure : Study selection 799 excluded after review of title and abstract 4 excluded after review of article 5 had no information on outcome of interest 8 had no original data 7 were in abstract form only See Online for appendix Vol 82 October 2, 269

3 consensus. As dictated by the Cochrane method, low risk of bias occurs when all domains are rated low, and means that bias is unlikely to seriously affect the results. An unclear risk of bias occurs when at least one domain has an unclear risk, and suggests that a bias that raises some doubts about the results might be present. Studies are rated at high risk for bias when at least one domain is rated high risk, and infers that the bias seriously weakens confidence in the results. Those of us who did the systematic review were not masked to authors, institution, or journal of publication. The use of non-masked reviewers is accepted in meta-analyses and has not been shown to bias results. Data synthesis We combined our studies using the DerSimonian and Laird random effects model. 29 We combined dichotomous outcomes (eg, all-cause mortality) using RRs, although odds ratios were examined in sensitivity analyses. We pooled continuous outcomes (eg, vascular calcification) with the weighted mean difference. We assessed heterogeneity among studies using the I² statistic, judging values of less than 25% to be minimal, less than 5% to be moderate, and 5% or greater to be substantial. We assessed publication bias both visually using a funnel plot and statistically using Egger s weighted regression and Duval s trim and fill method. We entered the data twice and compared each entry for all analyses to ensure that no data entry errors occurred. Our primary analyses assessed data from randomised controlled trials only. We did secondary analyses of the primary outcome measure (all-cause mortality) including non-randomised trials, by duration of follow-up, dialysis status, and type of non-calcium-based phosphate binder. Note that we used all available data reported in each study. As such, no data were missing. Role of the funding source This study had no funding source. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. Design Dialysis status of patients Mean age (years) Sex (% female) Diabetes (%) Mean dialysis duration (months) Non-calcium-based phosphate binder (n) Calcium-based phosphate binder (n) Follow-up (months) Original studies Chertow et al () 5 RCT Haemodialysis 57 5% 2 5% 9 Sevelamer (99) Calcium acetate or 2 Unclear calcium carbonate () Sadek et al (2) 6 RCT Haemodialysis Sevelamer (5) Calcium carbonate (6) 5 Unclear Braun et al (24) 7 RCT Haemodialysis % 8 8% 6 4 Sevelamer (56) Calcium carbonate (57) 2 High Block et al (25 8 and 9 ) RCT Haemodialysis 58 7% 6 6% 2 9 Sevelamer (54) Calcium acetate or 8, 44* Low calcium carbonate (55) Russo et al () RCT Non-dialysis 55 7% Sevelamer () Calcium carbonate () 24 High Borzecki et al () RCS Haemodialysis % 48 6% Sevelamer (68) Calcium acetate or 24 NA calcium carbonate (769) Barreto et al () 2 RCT Haemodialysis 47 2% 4 % Sevelamer (52) Calcium acetate (49) 2 Low Qunibi et al () RCT Haemodialysis 59 49% 57 % 2 Sevelamer () Calcium acetate () 2 Low Suki () 4 RCT Haemodialysis 6 46% 5 2% 8 2 Sevelamer (5) Calcium acetate or 2 High calcium carbonate (5) Takei et al () 5 RCT Haemodialysis 54 52% 5 7% Sevelamer () Calcium carbonate (2) 24 High New studies Wilson et al (29) 6 RCT Haemodialysis 54 7% 4 5% 4 2 Lanthanum (68) Calcium acetate or 24 High calcium carbonate (674) Panichi et al () 7 OS Haemodialysis 66 6% 6% Sevelamer (242) Calcium acetate or 24 NA calcium carbonate () Shantouf et al () 8 CS Haemodialysis 54 6 % 45 8% Sevelamer (57) Calcium carbonate or NA NA Calcium acetate (6) Jean et al () 9 OS Haemodialysis 68 4% 6% 62 6 Sevelamer (247) Calcium carbonate (42) 42 NA Kakuta et al () 2 RCT Haemodialysis 58 46% 2 8% 8 Sevelamer (9) Calcium carbonate (92) 2 Unclear Toussaint et al () RCT Haemodialysis % 7 8% Lanthanum () Calcium carbonate (2) 8 High Di Iorio et al (2) RCT Non-dialysis % % Sevelamer (7) Calcium carbonate (5) 6 Low Block et al (2) 2 RCT Non-dialysis % 57% Lanthanum (); sevelamer () Calcium acetate () 9 Low Our primary analyses included only RCTs. RCT=randomised controlled trial. RCS=retrospective cohort study. NA=not applicable. OS=observational study. CS=cross-sectional study. =data not reported. *8 month follow-up in the original study; 44 months median follow-up in the long-term outcome study. Table: Characteristics of studies included in the meta-analysis Risk of bias score Vol 82 October 2,

4 Results Our updated search identified 847 citations that included our keywords. Of these, 48 were retrieved for more detailed assessment, from which 4 were excluded: 5 because they did not provide information on the outcomes of interest, 8 because they did not present original data (review articles or commentaries), and seven because data was available only in abstract form. As such, we added eight new studies to the ten previously identified and reported studies to our updated meta-analysis (figure ). The newly selected eight studies included 2 participants, of whom 726 received calcium-based phosphate binders (calcium carbonate or calcium acetate) and 54 received non-calcium-based binders (774 received sevelamer and 7 received lanthanum). Of the trials that randomly assigned patients to sevelamer, three,, did not specify type (hydrochloride or carbonate), one 2 used sevelamer carbonate, and the remainder randomly assigned particpants to sevelamer hydrochloride. Study size ranged from 45 to 54 patients with follow-up from 9 to 6 months. Two studies,2 included patients with chronic kidney disease stages and 4 not on dialysis; the remainder included patients on haemodialysis. Of the eight newly added studies, two 7,9 were observational cohort studies, one 8 was a cross-sectional study, one 2 was a blinded randomised controlled trial, and the remaining four 6,2 were open-label randomised controlled trials. All the randomised controlled trials included in our original meta-analysis were open label. Risk of bias was rated as high in two randomised controlled trials, 6, one because of large losses to follow-up and the second because of a shortage of documentation in key tool domains. 6 Of all 8 studies included in the meta-analysis, all but four,7 9 reported serum phosphate concentrations at baseline and study conclusion, and no significant differences in the change in serum phosphate concentration by treatment assignment were reported. The table lists the key characteristics of the studies included in our original meta-analysis, together with the eight new studies. The results from our updated meta-analysis are shown in figures 2,, 4, and 5. The primary outcome, all-cause mortality, was based on the randomised trials that reported an outcome of mortality, and consisted of 46 patients with 96 deaths. 5,6,9,,2, 6,2, We noted that Non-calcium binders Events Total patients RCTs Barreto et al () 2 Block et al () 9 Chertow et al () 5 Di Iorio et al (2) Kakuta et al () 2 Qunibi et al () Russo et al () Sadek et al (2) 6 Suki () 4 Takei et al () 5 Wilson et al (29) Calcium binders Weight Events Total patients % 2% % % 8% % 24 5% 7 9% 5 9% Heterogeneity: τ²= ; χ²=2 5, df=7 (p= 9); I²=4% Test for overall effect: Z=2 9 (p= 4) Non-randomised studies Borzecki et al () % Jean et al () % Panichi et al () % % Heterogeneity: τ²= ; χ²= 57, df=2 (p= 46); I²=% Test for overall effect: Z= 9 (p= 6) Total % Heterogeneity: τ²= ; χ²= 88, df= (p= 8); I²=% Test for overall effect: Z=2 4 (p= 2) Test for subgroup differences: χ²= 92, df= (p= 4), I²=% Risk ratio (95% CI) 2 ( 2 9) 5 ( ) ( 9 88) 54 ( ) 44 ( 2 66) ( ) 97 ( 84 2) 85 ( 7 5) 78 ( 6 98) 82 ( 69 98) 99 ( 76 ) 9 ( 74 6) 89 ( 78 9) 87 ( 77 97) 2 5 Favours non-calcium Favours calcium Figure 2: All-cause mortality for each type of phosphate binder, according to type of trial All-cause mortality by type of phosphate binder, considering all randomised trials that reported an outcome of mortality, non-randomised trials that reported an outcome of mortality, and all trials together. Number of all mortality events and the total number of patients randomly assigned are presented for each treatment group. RC T=randomised controlled trial. Vol 82 October 2,

5 patients randomly assigned to non-calcium-based phosphate binders had a statistically significant % reduction in all-cause mortality compared with those randomly assigned to calcium-based phosphate binders (RR 78, 95% CI 6 98). Similarly, in the three non-randomised trials ( patients with 79 events), the reduction in all-cause mortality was % (RR 89, 78 ) in those taking non- calcium-based phosphate binders,7,9 and considering randomised and non-randomised trials together, the reduction in all-cause mortality was % (RR 87, 77 97) in favour of non-calcium-based phosphate binders (figure 2). In our analysis of differences in mortality by type of non-calcium-based binder, considering randomised and non-randomised trials together, we identified 2 studies 5,6,9 5,7,2, (ten of which were randomised Non-calcium binders Events Total patients Calcium binders Weight Events Total patients Risk ratio (95% CI) 6 months Sadek et al (2) 6 Heterogeneity: NA Test for overall effect: Z= 99 (p= 2) % % ( ) ( ) 2 months Barreto et al () 2 Chertow et al () 5 Kakuta et al () 2 Qunibi et al () Heterogeneity: τ 2 = 62; χ 2 =4, df=2 (p= 2); I 2 =5% Test for overall effect: Z= (p= 26) % % 8% 2 % 2 ( 2 9) ( 9 88) 44 ( 2 66) 49 ( 4 68) 8 months Block et al () 9 Suki () Heterogeneity: τ 2 = 2; χ 2 =, df= (p= 7); I 2 =69% Test for overall effect: Z= 87 (p= 9) % 24 5% 6% 5 ( ) 97 ( 84 2) 78 ( 45 7) 24 months Borzecki et al () Panichi et al () 7 Russo et al () Takei et al () 5 Wilson et al (29) Heterogeneity: τ 2 = ; χ 2 = 68, df=2 (p= 7); I 2 =% Test for overall effect: Z=2 62 (p= 9) % 5 9% 7 9% 54 % 82 ( 69 98) 9 ( 74 6) 85 ( 7 5) 86 ( 76 96) 6 42 months Di lorio et al (2) Jean et al () Heterogeneity: τ 2 = ; χ 2 =, df= (p= 8); I 2 =67% Test for overall effect: Z= 8 (p= 42) % 2 7% 5 7% 54 ( ) 99 ( 76 ) 79 ( 4 42) Total Heterogeneity: τ 2 = ; χ 2 = 88, df= (p= 8); I 2 =% Test for overall effect: Z=2 4 (p= 2) Test for subgroup differences χ 2 = 64, df=4 (p= 8), I 2 =% 426 % 87 ( 77 97) 2 5 Favours non-calcium Favours calcium Figure : All-cause mortality for each type of phosphate binder, according to length of follow-up All-cause mortality by length of follow-up, considering all trials that reported an outcome of mortality. Number of all mortality events and the total number of patients randomly assigned are presented for each treatment group. NA=not applicable. 2 Vol 82 October 2,

6 Non-calcium binders Events Total patients Calcium binders Weight Events Total patients Risk ratio (95% CI) Dialysis Barreto et al () 2 Block et al () 9 Borzecki et al () Chertow et al () 5 Jean et al () 9 Kakuta et al () 2 Panichi et al () 7 Qunibi et al () Sadek et al (2) 6 Suki () 4 Takei et al () 5 Wilson et al (29) 6 (95% CI) Heterogeneity: τ 2 = ; χ 2 = 52, df=9 (p= 24); I 2 =% Test for overall effect: Z=2 2 (p= ) % 2% 2 6% % 2 7% 5 9% 8% % 24 5% 7 9% 97 % 2 ( 2 9) 5 ( ) 82 ( 69 98) ( 9 88) 99 ( 76 ) 9 ( 74 6) 44 ( 2 66) ( ) 97 ( 84 2) 85 ( 7 5) 88 ( 79 99) Predialysis Di lorio et al (2) 2 Russo et al () 2 Heterogeneity: NA Test for overall effect: Z= 89 (p= 6) % % 54 ( ) 54 ( ) Total Heterogeneity: τ 2 = ; χ 2 = 88, df= (p= 8); I 2 =% Test for overall effect: Z=2 4 (p= 2) Test for subgroup differences χ 2 =2, df= (p= 4), I 2 =55 % 426 % 87 ( 77 97) 2 5 Favours non-calcium Favours calcium Figure 4: All-cause mortality for each type of phosphate binder, according to dialysis status All-cause mortality by dialysis status, considering all trials that reported an outcome of mortality. Number of all mortality events and the total number of patients randomly assigned are presented for each treatment group. trials) that reported on mortality in patients taking sevelamer (n=2669; 574 events) compared with calcium-based phosphate binders (n=5; 8 events) and one randomised trial 6 that reported on mortality in patients taking lanthanum (n=68; 5 events) compared with calcium-based phosphate binders (n=674; 57 events). We noted a non-statistically significant decrease in mortality in patients randomly assigned to sevelamer (RR 89, 95% CI 78 ) and lanthanum (RR 74, 49 ), compared with those randomly assigned to calcium-based phosphate binders. In our analyses of differences in mortality by duration of follow-up, we only noted a statistically significant difference in mortality between patients assigned to non-calcium-based binders and those assigned to calcium-based binders in the five trials that reported outcomes at 24 months; these trials had the largest number of patients and events (figure ). We also examined differences in mortality by dialysis status (ie, predialysis and dialysis) and noted that both in patients on dialysis and those not, mortality was reduced in patients assigned to non-calcium-based phosphate binders compared with those assigned to calcium-based binders (figure 4). None of the eight newly added studies included data for cardiovascular events, therefore our estimate of effect is the same as in our previous paper: a non-significant reduction in mortality of 5% (RR 85, 95% CI 5 2 ). 24 Coronary artery calcification was reported in seven randomised trials 5,9,,2,,2,2 (five of these were included in our previously published meta-analysis). Data for coronary artery calcification was reported in 74 patients; 25 patients were newly added. All these trials used the Agatston score to rate coronary artery calcification, and in all but two studies 2,2 scans were read by an assessor masked to the treatment assignment. We organised the data by length of follow-up, starting with a minimum interscan time of 6 months. The reduction in vascular calcification was greater in patients assigned to non-calcium- based Vol 82 October 2,

7 phosphate binders than in those assigned to calcium binders at all timepoints; this finding was statistically significant when we analysed data from the longest follow-up point for each study (mean difference in Agatston score 95 26, 95% CI to 4 84; figure 5). Non-calcium binders Calcium binders Weight Mean SD Total patients Mean 6 month follow-up Block et al (25) 8 Block et al (2) 2 Chertow et al () 5 Qunibi et al () (95% CI) Heterogeneity: τ 2 =475 24; χ 2 =4 95, df= (p= 8); I 2 =9% Test for overall effect: Z= (p= ) SD 452 Total patients 5 6% % 7% 42 % 4 % Mean difference (95% CI) 2 ( 76 8 to 2 8) 98 ( 47 to 549 7) 244 ( 46 9 to 5 6) 2 ( 2 4 to 88 4) 62 2 ( 7 7 to 47 2) 2 month follow-up Barreto et al () 2 Block et al (25) 8 Braun et al (24) 7 Chertow et al () 5 Kakuta et al () 2 Qunibi et al () (95% CI) Heterogeneity: τ 2 =464 67; χ 2 =5 49, df=5 (p= 6); I 2 =9% Test for overall effect: Z= 44 (p= 6) % 6 % % 6 8% 47 6% 2 7% % 4 ( to 96 99) 82 ( 87 to 47 87) ( to 5 57) 97 ( 4 56 to 7 56) 2 2 ( 79 to 45 4) ( 48 9 to 46 9) ( 5 4 to 4 29) 8 month follow-up Block et al (25) 8 (95% CI) Heterogeneity: NA 8 42 Test for overall effect: Z= 6 (p= ) % % 2 ( to 42 84) 2 ( to 42 84) 24 month follow-up Russo et al () (95% CI) Heterogeneity: NA 8 77 Test for overall effect: Z= 62 (p= 54) 7 % % 95 ( 96 7 to 26 77) 95 ( to 26 77) All studies: longest timepoint Barreto et al () 2 Block et al (25) 8 Block et al (2) 2 Chertow et al () 5 Kakuta et al () 2 Qunibi et al () Russo et al () (95% CI) Heterogeneity: τ 2 = ; χ 2 =4 76, df=6 (p= 57); I 2 =% Test for overall effect: Z= 6 (p= ) % 4 5% % 6 % 59 % 2 2% 2 9% % 4 ( to 96 99) 2 ( to 42 84) 98 ( 47 to 549 7) 97 ( 4 56 to 7 56) 2 2 ( 79 to 45 4) ( 48 9 to 46 9) 95 ( to 26 77) ( to 4 84) Non-randomised studies Shantouf et al () 8 (95% CI) Heterogeneity: NA 8 Test for overall effect: Z=2 89 (p< ) % % ( 24 9 to 78 9) ( 24 9 to 78 9) Favours non-calcium Favours calcium Figure 5: Coronary artery calcification, according to length of follow-up Mean (SD) Agatston score and the total number of patients randomly assigned are presented for each treatment group. NA=not applicable. 4 Vol 82 October 2,

8 SE (log [risk ratio]) Risk ratio Figure 6: Funnel plot for all-cause mortality Vertical dotted line shows the average risk ratio across all randomised trials reporting an outcome of mortality. As in our previous meta-analysis, none of the studies reported on fractures or vascular compliance. The funnel plot to assess publication bias was somewhat asymmetric (figure 6) and this was confirmed with Egger s regression test (p= 2). This shows that our findings might be slightly overestimating the true effect size if we are missing non-significant studies; although Duval s trim and fill analysis added no new studies to the analysis. Discussion Our systematic review of randomised trials included 46 patients and 96 deaths and showed a % reduction in all-cause mortality in patients receiving non-calciumbased phosphate binders (sevelamer or lanthanum) compared with those assigned to calcium-based regimens. We noted similar reductions in mortality in non-randomised trials and when we considered predialysis and dialysis patients separately. Length of follow-up varied in each study; studies with the largest number of patients and outcomes followed up patients for 24 months. Follow-up at 24 months, but not 6 42 months, was associated with a significant reduction in mortality in patients randomly assigned to non-calcium-based versus calcium-based binders, which is probably because there was insufficient power to assess the outcome at 6 42 months (figure ). Note that the reduction in mortality associated with non-calcium-based phosphate binders seems to be independent of serum phosphate reduction, which did not differ by treatment assignment. In view of the burden of cardiovascular disease in this population of patients, we suggest that the first-line therapy for phosphate lowering should be non-calcium-based binders specifically sevelamer or lanthanum. Whether other non- calcium-based phosphate binders can be regarded as equally effective and safe remains to be shown. One potential explanation for the decrease in mortality associated with the use of non-calcium-based binders might be related to slowing of vascular calcification. We noted an increase in coronary artery calcification in patients taking calcium-based phosphate binders compared with those taking non-calcium-based binders, which is consistent with this explanation. Progression of coronary artery calcification has been associated with an increase in vascular stiffness and QT dispersion on the surface electrocardiogram, regarded as a marker of increased risk of ventricular arrhythmias. 2 Unfortunately, we were unable to show whether the increase in mortality in patients taking calcium-based binders was due to an increase in cardiovascular events, because this finding was not reported in any of the new studies in our updated meta-analysis. To show whether mortality differed by type of non-calcium-based phosphate binder, we examined the 2 trials that compared sevelamer with calcium-based phosphate binders and the two trials that examined effects of lanthanum. This a-priori subgroup analysis showed a statistically non-significant decrease in mortality in patients taking either sevelamer or lanthanum compared with those taking calcium-based phosphate binders. Clearly, further research is needed to unequivocally show the effects of calcium-based versus non-calcium-based phosphate binders on mortality, investigate the mechanisms of death in patients who are taking phosphate binders, and to show whether outcome differs by specific type of non-calcium-based binder used. Our systematic review has several strengths, including the broad search strategy with standard Cochrane protocols and the large study sample (we were able to abstract data from about 5 patients, of whom more than 9 had died). The studies identified were of reasonable quality with no evidence of differential dropout by treatment assignment. Limitations of our review include the fact that no new studies examined cardiovascular mortality and as such we cannot comment on differences in this outcome on the basis of type of binder used. Secondly, as has been suggested in other studies, we assumed that calcium carbonate and calcium acetate were equivalent. Thirdly, only one study reported specifically on the use of sevelamer carbonate, the rest randomly assigned patients to sevelamer (type not specified), sevelamer hydrochloride, or lanthanum. Although sevelamer and lanthanum are the most frequently prescribed non-calcium-based binders, we cannot comment on the effect of other non-calcium-based binders on mortality, nor can we comment on whether sevelamer carbonate differs from sevelamer hydrochloride, especially in patients with late-stage chronic kidney disease not yet on dialysis. Our funnel plot analysis showed a potential publication bias (figure 6), but this bias is always difficult to ascertain with such a small amount of studies. None of the studies formally assessed adverse events, and follow-up was Vol 82 October 2, 5

9 short although this would bias in favour of finding a decrease in mortality over time. As might be expected, a moderate degree of heterogeneity was noted between studies, and only a third of the studies were assessed as having a low risk of bias. Nevertheless, we felt it important to present the totality of the evidence. Finally, because of the design of the studies included in this review, we cannot establish whether non-calcium-based phosphate binders are inherently beneficial, or if calcium-based binders are harmful, or a bit of both. Patients with chronic kidney disease are at high risk of mortality. Phosphate binding and lowering of serum phosphate can be achieved with calcium-based or non-calcium-based binders. Calcium-based phosphate binders are inexpensive; however, based on our systematic review, the best evidence available so far suggests that they might be harmful. The mechanism of the benefits of non-calcium-based phosphate binders seems to be a slower progression of vascular calcification. This mechanism needs to be confirmed by careful studies of cardiovascular outcomes. Contributors SAJ selected and reviewed studies identified by the scientific literature search, did the Cochrane risk of bias, wrote drafts with RTT, and wrote the final paper. BV contributed to the statistical analysis and data synthesis of outcomes, and drafted and edited the final paper. PR reviewed pertinent scientific literature, helped with the first draft of the paper and subsequent versions, and provided critical insight into the interpretation and meta-analytical data. DCM was involved in the conception and design of the study, reviewed abstracts to select core articles, reviewed and abstracted data from core articles, reviewed and commented on each draft, and approved final submission. TC and MD searched the scientific literature, contributed to the writing of the methods section, and edited the paper. CEL designed the study and obtained, analysed, and interpreted the data. DF did the planning analysis, reviewed articles selected by TC and MD, and extracted data from articles. RTT advised on the study methods, reviewed and selected the studies, extracted the data, and wrote drafts of the manuscript. Conflicts of interest PR has received honoraria for speaking engagements for Sanofi. DCM is on advisory boards for Genzyme and Amgen, and has received a speaking honorarium from Amgen. CEL is a consultant for Gore, Roche, and Takeda. RTT has received consulting fees from Genzyme. All other authors declare that they have no conflicts of interest. References Levey AS, Coresh J. Chronic kidney disease. Lancet 2; 79: Block GA, Hulbert-Shearon TE, Levin NW, Port FK. Association of serum phosphorus and calcium x phosphate product with mortality risk in chronic hemodialysis patients: a national study. Am J Kidney Dis 998; : Jono S, McKee MD, Murry CE, et al. Phosphate regulation of vascular smooth muscle cell calcification. Circ Res 2; 87: E 7. 4 Molony DA, Stephens BW. Derangements in phosphate metabolism in chronic kidney diseases/endstage renal disease: therapeutic considerations. Adv Chronic Kidney Dis ; 8: 2. 5 Chertow GM, Burke SK, Raggi P. Sevelamer attenuates the progression of coronary and aortic calcification in hemodialysis patients. Kidney Int ; 62: Sadek T, Mazouz H, Bahloul H, et al. Sevelamer hydrochloride with or without alphacalcidol or higher dialysate calcium vs calcium carbonate in dialysis patients: an open-label, randomized study. Nephrol Dial Transplant 2; 8: Braun J, Asmus HG, Holzer H, et al. Long-term comparison of a calcium-free phosphate binder and calcium carbonate phosphorus metabolism and cardiovascular calcification. Clin Nephrol 24; 62: Block GA, Spiegel DM, Ehrlich J, et al. Effects of sevelamer and calcium on coronary artery calcification in patients new to hemodialysis. Kidney Int 25; 68: Block GA, Raggi P, Bellasi A, Kooienga L, Spiegel DM. Mortality effect of coronary calcification and phosphate binder choice in incident hemodialysis patients. Kidney Int ; 7: Russo D, Miranda I, Ruocco C, et al. The progression of coronary artery calcification in predialysis patients on calcium carbonate or sevelamer. Kidney Int ; 72: Borzecki AM, Lee A, Wang SW, Brenner L, Kazis LE. Survival in end stage renal disease: calcium carbonate vs sevelamer. J Clin Pharm Ther ; 2: Barreto DV, Barreto FdC, de Carvalho AB, et al. Phosphate binder impact on bone remodeling and coronary calcification results from the BRiC study. Nephron ; : c 8. Qunibi W, Moustafa M, Muenz LR, et al. A -year randomized trial of calcium acetate versus sevelamer on progression of coronary artery calcification in hemodialysis patients with comparable lipid control: the Calcium Acetate Renagel Evaluation-2 (CARE-2) study. Am J Kidney Dis ; 5: Suki WN, for the Dialysis Clinical Outcomes Revisited Investigators. Effects of sevelamer and calcium-based phosphate binders on mortality in hemodialysis patients: results of a randomized clinical trial. J Ren Nutr ; 8: Takei T, Otsubo S, Uchida K, et al. Effects of sevelamer on the progression of vascular calcification in patients on chronic haemodialysis. Nephron ; 8: c Wilson R, Zhang P, Smyth M, Pratt R. Assessment of survival in a 2-year comparative study of lanthanum carbonate versus standard therapy. Curr Med Res Opin 29; 25:. 7 Panichi V, Bigazzi R, Paoletti S, et al. Impact of calcium, phosphate, PTH abnormalities and management on mortality in hemodialysis: results from the RISCAVID study. J Nephrol ; 2: Shantouf R, Ahmadi N, Flores F, et al. Impact of phosphate binder type on coronary artery calcification in hemodialysis patients. Clin Nephrol ; 74: Jean G, Lataillade D, Genet L, et al. Calcium carbonate, but not sevelamer, is associated with better outcomes in hemodialysis patients: results from the French ARNOS study. Hemodial Int ; 5: Kakuta T, Tanaka R, Hyodo T, et al. Effect of sevelamer and calcium-based phosphate binders on coronary artery calcification and accumulation of circulating advanced glycation end products in hemodialysis patients. Am J Kidney Dis ; 57: 4. Toussaint ND, Lau KK, Polkinghorne KR, Kerr PG. Attenuation of aortic calcification with lanthanum carbonate versus calcium-based phosphate binders in haemodialysis: a pilot randomized controlled trial. Nephrology (Carlton) ; 6: Di Iorio B, Bellasi A, Russo D. Mortality in kidney disease patients treated with phosphate binders: a randomized study. Clin J Am Soc Nephrol 2; 7: Block GA, Wheeler DC, Persky MS, et al. Effects of phosphate binders in moderate CKD. J Am Soc Nephrol 2; 2: Jamal SA, Fitchett D, Lok CE, Mendelssohn DC, Tsuyuki RT. The effects of calcium-based versus non-calcium-based phosphate binders on mortality among patients with chronic kidney disease: a meta-analysis. Nephrol Dial Transplant 29; 24: Moher D, Liberati A, Tetzlaff J, Altman DG, for the PRISMA group. Preferred reporting items for systematic reviews and meta-analyses; the PRISMA statement. Ann Intern Med 29; 5: van Tulder M, Furlan A, Bombardier C, Bouter L. Updated method guidelines for systematic reviews in the cochrane collaboration back review group. Spine 2; : Higgins JPT, Green S, eds. Cochrane handbook for systematic reviews of interventions, version Chichester, West Sussex: The Cochrane Collaboration and John Wiley & Sons,. Berlin JA, for the University of Pennsylvania Meta-analysis Blinding Study Group. Does blinding of readers affect the results of meta-analyses? Lancet 997; 5: Vol 82 October 2,

10 29 DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 986; 7: Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ 997; 5: Duval S, Tweedie R. Trim and fill: a simple funnel-plot-based method of testing and adjusting for publication bias in meta-analysis. Biometrics 2; 56: Di Iorio B, Nargi O, Cucciniello E, et al. Coronary artery calcification progression is associated with arterial stiffness and cardiac repolarization deterioration in hemodialysis patients. Kidney Blood Press Res ; 4: Chertow GM, Raggi P, McCarthy JT, et al. The effects of sevelamer and calcium acetate on proxies of atherosclerotic and arteriosclerotic vascular disease in hemodialysis patients. Am J Nephrol 2; 2: Vol 82 October 2, 7