Effect of B-Vitamin Therapy on Progression of Diabetic Nephropathy
A Randomized Controlled Trial
JAMA. May 2010;303(16):1603-1609.|
Andrew A. House, MD; Misha Eliasziw, PhD; Daniel C. Cattran, MD; David N. Churchill, MD; Matthew J. Oliver, MD; Adrian Fine, MD; George K. Dresser, MD; J. David Spence, MD
Author Affiliations: Division of Nephrology (Dr House), Division of Clinical Pharmacology (Drs Dresser and Spence), and Robarts Research Institute (Dr Spence), University of Western Ontario, London, Ontario; Department of Community Health Sciences, Clinical Neurosciences, and Oncology, University of Calgary, Calgary, Alberta (Dr Eliasziw); Division of Nephrology, University of Toronto, Toronto, Ontario (Drs Cattran and Oliver); Division of Nephrology, McMaster University, Hamilton, Ontario (Dr Churchill); and Division of Nephrology, University of Manitoba, Winnipeg, Manitoba (Dr Fine), Canada.
Context Hyperhomocysteinemia is frequently observed in patients with diabetic nephropathy. B-vitamin therapy (folic acid, vitamin B6, and vitamin B12) has been shown to lower the plasma concentration of homocysteine.
Objective To determine whether B-vitamin therapy can slow progression of diabetic nephropathy and prevent vascular complications.
Design, Setting, and Participants A multicenter, randomized, double-blind, placebo-controlled trial (Diabetic Intervention with Vitamins to Improve Nephropathy [DIVINe]) at 5 university medical centers in Canada conducted between May 2001 and July 2007 of 238 participants who had type 1 or 2 diabetes and a clinical diagnosis of diabetic nephropathy.
Intervention Single tablet of B vitamins containing folic acid (2.5 mg/d), vitamin B6 (25 mg/d), and vitamin B12 (1 mg/d), or matching placebo.
Main Outcome Measures Change in radionuclide glomerular filtration rate (GFR) between baseline and 36 months. Secondary outcomes were dialysis and a composite of myocardial infarction, stroke, revascularization, and all-cause mortality. Plasma total homocysteine was also measured.
Results The mean (SD) follow-up during the trial was 31.9 (14.4) months. At 36 months, radionuclide GFR decreased by a mean (SE) of 16.5 (1.7) mL/min/1.73 m2 in the B-vitamin group compared with 10.7 (1.7) mL/min/1.73 m2 in the placebo group (mean difference, -5.8; 95% confidence interval [CI], -10.6 to -1.1; P = .02). There was no difference in requirement of dialysis (hazard ratio [HR], 1.1; 95% CI, 0.4-2.6; P = .88). The composite outcome occurred more often in the B-vitamin group (HR, 2.0; 95% CI, 1.0-4.0; P = .04). Plasma total homocysteine decreased by a mean (SE) of 2.2 (0.4) µmol/L at 36 months in the B-vitamin group compared with a mean (SE) increase of 2.6 (0.4) µmol/L in the placebo group (mean difference, -4.8; 95% CI, -6.1 to -3.7; P < .001, in favor of B vitamins).
Conclusion Among patients with diabetic nephropathy, high doses of B vitamins compared with placebo resulted in a greater decrease in GFR and an increase in vascular events.
Diabetic nephropathy is an important cause of chronic kidney disease.1-2 In the United States, diabetes mellitus accounts for more than 44% of patients with end stage renal failure.3 In addition to the personal burden, the societal burden of diabetic nephropathy is enormous, exceeding US $10 billion in annual medical expenditures.4 Despite effective therapies to slow disease progression,5 approximately 40% of the estimated 21 million patients with diabetes in the United States develop overt nephropathy.4 New treatment approaches to this problem are needed.
Several observational studies have shown a significant association between high concentrations of plasma total homocysteine and the risk of developing diabetic nephropathy, retinopathy, and vascular diseases, including myocardial infarction (MI) and stroke.6-9 B-vitamin therapy has been shown to lower the plasma concentration of homocysteine and improve endothelial function.10-11
We hypothesized that B-vitamin therapy would slow the progression of diabetic nephropathy and prevent vascular events. To test this hypothesis, we conducted a multicenter, randomized, double-blind, placebo-controlled trial of participants with type 1 or 2 diabetes.
In this multicenter, randomized, double-blind, placebo-controlled trial, we found that high doses of combined B vitamins (folic acid, vitamin B6, and vitamin B12) significantly lowered plasma total homocysteine in participants with diabetic nephropathy compared with matching placebo. However, the group of participants assigned to the B-vitamin therapy had a more rapid decrease in renal function, measured by radionuclide GFR, and had a higher rate of MI and stroke.
There is a strong biological rationale for homocysteine to be a risk factor for vascular disease, as shown in animal models,19-20 in vitro scientific work,21-23 and observational studies.6-7,24 However, most large homocysteine-lowering B-vitamin clinical trials have failed to show a benefit for the prevention of stroke17 and cardiovascular outcomes.25-28 One interpretation of the conflicting results is that the mechanisms by which homocysteine influences vascular outcomes are not primarily related to the pathogenesis of coronary artery disease, which is largely due to plaque formation and rupture, although an imaging substudy showed that patients receiving B-vitamin therapy had less plaque progression of 3-dimensional vessel wall volume using ultrasound.29 Any putative role of high homocysteine levels might, therefore, be at the level of the microvasculature promoting endothelial dysfunction and lacunar infarction, and through increased thrombosis.30
Our trial is the first study to our knowledge to show significant detrimental effects from pharmacological doses of B vitamins (folic acid, vitamin B6, and vitamin B12). Other studies have suggested potential harm, or at best a neutral effect of B vitamins. Our results are consistent with a recent small study31 showing increased vascular events in patients with unstable angina and non-ST-elevation MI, particularly in the subgroup with diabetes. Another recent publication has raised concerns about increased cancer risk and all-cause mortality in patients with ischemic heart disease receiving B vitamins.32 Although our study was limited by its sample size to detect differences in cancer and other infrequent outcomes, it was adequately powered to test our hypothesis regarding renal function decline and did show a significant difference in predefined cardiovascular and cerebrovascular events. Whether these results can be generalized to other populations beyond those similar to the participants enrolled into the trial or dosing strategies that differ from the combination chosen herein is speculative. Patient discontinuation over time did not affect our results, because there were no differences in baseline characteristics between participants who were and were not available for the primary 36-month analyses.
One interpretation of our completed trial is that these results are simply the play of chance. This explanation seems unlikely, given that renal function decline was mirrored by the occurrence of vascular outcomes, suggesting that B vitamins were in some way associated with both renal and vascular toxicity. An alternative explanation is that homocysteine lowering may be protective, but offset by toxicity associated with pharmacological doses of B vitamins. At the beginning of the trial, serum levels of B vitamins were rather high among participants randomized to the B-vitamin group and by the end of the trial, they were very high. Mandatory folic acid fortification likely contributed to the observation of relatively high folate levels in all participants, limiting generalizability of our results to an unsupplemented population. Because these vitamins are water-soluble and renally excreted, vitamin toxicity may be more of a concern in patients with impaired renal function. One or more of 3 potential mechanisms33 may explain the renal and vascular toxicity observed in our trial. First, folic acid may promote cell proliferation through its role in thymidine synthesis. Second, the use of folic acid and vitamin B12 might alter the methylation potential in vascular cells. Finally, B-vitamin therapy could potentially increase the methylation of L-arginine to the nitric oxide synthase inhibitor asymmetric dimethylarginine.
If the homocysteine theory of atherosclerosis is to be proved or disproved with certainty, it may be necessary to find alternative, nonvitamin strategies to lower homocysteine, such as enhancing the conversion of homocysteine to cysteine in the liver or enhancing urinary excretion.33 To this end, investigators have begun to examine the use of agents capable of thiol exchange, liberating homocysteine bound by disulfide bonds to proteins; therefore, facilitating its urinary excretion and lowering plasma levels significantly without the use of vitamins.34
In conclusion, in our randomized trial of participants with diabetic nephropathy and stages 1 to 3 chronic kidney disease, the use of high doses of B vitamins (containing 2.5 mg/d of folic acid, 25 mg/d of vitamin B6, and 1 mg/d of vitamin B12) compared with placebo resulted in a greater decrease in GFR and an increase in MI and stroke. Given the recent large-scale clinical trials showing no treatment benefit, and our trial demonstrating harm, it would be prudent to discourage the use of high-dose B vitamins as a homocysteine-lowering strategy outside the framework of properly conducted clinical research.
Figure 1 shows the flow of participants in the trial. Four participants did not meet inclusion criteria and 10 did not return for their first postrandomization visit to provide their GFR efficacy measure. Therefore, 238 participants (119 in the B-vitamin group and 119 in the placebo group) were available for the modified intention-to-treat analyses. Participants were followed up for a mean (SD) of 31.9 (14.4) months and a median (interquartile range) of 32.2 (28.2) months.
Baseline characteristics were similar between the placebo and B-vitamin groups (Table 1). Participants were predominantly men (178 [74.8%]) and white (198 [83.2%]), with type 2 diabetes (195 [81.9%]). Consistent with the presence of diabetic nephropathy, the mean (SD) baseline GFR was low at 54.7 (29.5) mL /min/1.73 m2, with 152 participants (63.9%) having at least stage 3 chronic kidney disease (GFR <60 mL /min/1.73 m2). The baseline mean plasma total homocysteine was increased at 15.5 (5.2) µmol/L. The prevalent use of hypoglycemic medications, insulin, angiotensin blockade, and other cardioprotective medications indicated that participants were well-treated by their attending physicians.
Among the 238 participants at baseline, 205 and 118 remained in the analyses at 18 and 36 months, respectively. Only 25 (12.2%) and 16 (13.6%) GFRs were unavailable at 18 and 36 months, respectively, and had to be imputed. Participants assigned to the B-vitamin group had a much greater decrease in radionuclide GFR compared with the placebo group (P = .045) (Table 2). Over 36 months, the GFR decreased by a mean (SE) of 16.5 (1.7) mL/min/1.73 m2 in the B-vitamin group compared with 10.7 (1.7) mL/min/1.73 m2 in the placebo group, a significant mean difference of -5.8 (95% CI, -10.6 to -1.1; P = .02). This result was similar to the MDRD formula results, which had estimated GFR measurements at 12 and 24 months.
Despite the unexpected reverse findings for GFR, the plasma total homocysteine results were as expected. At 36 months, participants in the B-vitamin group had a mean (SE) decrease of 2.2 (0.4) µmol/L and participants in the placebo group had a mean (SE) increase of 2.6 (0.4) µmol/L, resulting in a significant mean difference of -4.8 (95% CI, -6.1 to -3.7; P < .001, in favor of B vitamins). This result was congruent with the significantly large increases (P < .001) in mean serum folate and serum vitamin B12 at all time points in the B-vitamin group (eTable). Proteinuria did not significantly change during the follow-up period (Table 2), and there was no difference in the proportion of participants who required dialysis (HR, 1.1; 95% CI, 0.4-2.6; P = .88) (Table 3).
Participants randomized to receive B vitamins had a significantly greater number of cardiovascular and cerebrovascular events (Table 3). The 36-month risk of a composite outcome, including MI, stroke, revascularization, and all-cause mortality, in the B-vitamin group was double that in the placebo group (HR, 2.0; 95% CI, 1.0-4.0; P = .04) (Table 3), and the corresponding cumulative failure curves are shown in Figure 2. The 36-month risks of the composite end point was 23.5% (95% CI, 15.0%-32.0%) in the B-vitamin group and 14.4% (95% CI, 6.9%-21.8%) in the placebo group. All-cause mortality and amputation (Table 3) and cognitive decline based on the Mini-Mental State Examination score (Table 2) did not significantly differ between the treatment groups. There were no significant differences in mean glycated hemoglobin between the 2 groups during the course of the trial (eTable). Mean systolic and diastolic blood pressures significantly decreased over time but did not differ between treatment groups, and most elements of the lipid profile also did not differ between treatment groups (eTable).
Adverse Events Other Than Outcomes
A total of 1060 adverse events were reported (535 in the placebo group and 525 in the B-vitamin group). These included many transient minor complaints, such as dizziness, nausea, or headache. None was deemed by the attending physician to be related to trial medication. Counting participants only once, 1 or more adverse events were reported in 107 participants (89.9%) in the placebo group and 105 (88.2%) in the B-vitamin group (P = .68). One or more severe adverse events were reported in 38 participants (31.9%) in the placebo group and 39 (32.8%) in the B-vitamin group (P = .89). One or more serious adverse events were reported in 48 participants (40.3%) in the placebo group and 40 (33.6%) in the B-vitamin group (P = .28).
The Diabetic Intervention with Vitamins to Improve Nephropathy (DIVINe) trial was conducted between May 2001 and July 2007 at 5 university medical centers in Canada. Participants were recruited from nephrology and diabetes clinics. Eligible participants had type 1 or 2 diabetes and a clinical diagnosis of diabetic nephropathy, with at least 300 mg/d of urinary albumin excretion (or 500 mg/d of proteinuria). All participants were 18 years or older. Race/ethnicity was self-reported at the time of enrollment and included because of a very high prevalence of diabetic nephropathy among Canadian aboriginal people. Potential participants who were expected to survive fewer than 3 years, those with advanced renal failure (defined as stage 4 or 5 chronic kidney disease12 with creatinine clearance of <30 mL/min13 or on dialysis), those awaiting imminent dialysis, and women who were pregnant or unwilling to practice effective contraception were not eligible for the trial. The trial protocol was approved by the institutional review board at each center and all participants gave written informed consent.
Assessment, Randomization, and Follow-up
Baseline assessments for trial participants included medical history, laboratory analyses, and current medications. Participants were randomized 1:1 to receive a single tablet of B vitamins that contained 2.5 mg/d of folic acid, 25 mg/d of vitamin B6, and 1 mg/d of vitamin B12, or a matching placebo. A computer-generated randomization list was generated by the trial biostatistician (M.E.) using permuted block sizes of 2 and 4, and was stratified by center, sex, and baseline Cockcroft-Gault formula creatinine clearance (<50 vs 50 mL /min/1.73 m2; to convert to mL /s/m2, multiply by 0.0167).
B vitamins and matching placebos were bottled with participant numbers by pharmacy staff. As trial participants were given the next appropriately labeled bottle, all participants, research coordinators, and treating physicians remained blinded to the treatment assignment throughout the trial.
Participants were followed up with clinic visits every 6 months for up to 3 years. Participants were permitted to take usual doses of vitamins E, C, and beta carotene, and multivitamins containing usual low doses of B vitamins, but were asked not to take additional doses of the trial B vitamins. All other treatments were left to the discretion of the participants' attending physicians. Adherence to therapy was assessed by pill counts at every clinic visit and by annual measurements of plasma total homocysteine, serum folate, and serum B12. Central laboratory reports were made available to attending physicians, and blinding was maintained by deleting the values for plasma total homocysteine, serum folate, and serum B12. Plasma total homocysteine was measured by Liquid Chromatography Tandem Mass Spectrometry (MDS Laboratories Inc, Toronto, Ontario, Canada). The same laboratory measured serum folate and serum B12, as well as serum glycated hemoglobin, serum creatinine, and urinary protein.
The primary outcome measure was progression of nephropathy, which was assessed by change in glomerular filtration rate (GFR). A validated technique for measuring radionuclide GFR was performed at baseline, 18 months, and 36 months using 99Technetium-DTPA (99Tc-diethylenetriaminepentaacetic acid).14 Glomerular filtration rate was also estimated every 6 months using a 3-hour timed creatinine clearance commencing 1 hour after a single dose of 800-mg cimetidine (to prevent tubular secretion of creatinine).14-15 Three values for each method were obtained at 30-minute intervals per participant visit. All radionuclide GFRs and cimetidine creatinine clearances were adjudicated by a blinded nephrologist (A.A.H.). In cases of outlying values (ie, >10% difference among the 3 values), the nephrologist discarded those values believed to be biologically implausible, most commonly based on inadequate urine flow. For each method, the median of the 3 values was used in the analyses. Glomerular filtration rate was also estimated by the 4-variable Modification of Diet in Renal Disease (MDRD) formula.16
Secondary and tertiary outcomes included dialysis, occurrence of vascular events and all-cause mortality, cognitive decline (measured by the Mini-Mental State Examination score), and amputation. Ascertainment of possible events was sought during participants' clinic visits and from hospital records. A blinded adjudication committee reviewed all events to determine their validity and to classify the vascular events as MI, stroke, revascularization (peripheral, cardiac angioplasty, or cardiac bypass), cause of death, and amputation for peripheral vascular disease, using the definitions of the Vitamin Intervention for Stroke Prevention trial.17 In addition, all other undesirable or harmful changes in health that occurred among participants in the trial were recorded and classified as adverse events. Adverse events that were incapacitating with inability to perform usual activity were classified as severe. Adverse events that caused death, permanent damage, birth defects, were life-threatening, or required hospitalization were classified as serious.
Sample Size and Statistical Analyses
A previous study had reported a mean (SD) decrease in GFR of 4.4 (2.9) mL/min/1.73 m2 per year.18 Given an observed constant rate of decline over time, we assumed a 3-year mean (SD) decrease of 12 (9) mL/min/1.73 m2 in the placebo group. A total trial size of 286 participants provided 80% power to detect a 25% reduction in GFR decline between the placebo and B-vitamin groups (12 vs 9 mL/min/ 1.73 m2), at a 2-tailed 5% level of significance. Interim efficacy analyses were not planned.
During the trial, the data and safety monitoring board committee noted that, although the rate of GFR decrease in the placebo group was comparable with that predicted, the rate of GFR decrease in the B-vitamin group was higher than the predicted placebo rate. The data and safety monitoring board committee performed a conditional power calculation that indicated improbability of yielding a significant benefit in the primary outcome at the end of the trial, and therefore recommended stopping enrollment after 252 participants and stopping follow-up after the final radionuclide GFR was completed.
Trial data were collected by the center coordinators using standard 2-part carbonless copy paper, monitored for completeness and accuracy by the study monitor, and double-keyed into an Oracle Clinical database at Robarts Clinical Trials, London, Ontario, Canada. Biochemical results from the central laboratory were received electronically. Data were analyzed using a modified intention-to-treat approach, defined as participants who met all the inclusion criteria and none of the exclusion criteria, ingested their first tablet of medication, and had 1 or more postrandomization GFR efficacy measures. Statistical analyses were performed by using SAS software version 9.2 (SAS Institute Inc, Cary, North Carolina) and multiple imputation was performed by using SOLAS software version 3.0 (Statistical Solutions Ltd, Saugus, Massachusetts).
The primary analysis compared treatment groups on their mean change in radionuclide GFR between baseline and 36 months using a mixed linear model (PROC MIXED procedure in SAS [SAS Institute Inc]), analogous to an unbalanced repeated measures analysis of covariance. For the analysis, participant was considered as a random effect, treatment group and visit number as fixed effects, an interaction term between treatment and visit to assess for a differential treatment effect over time, and baseline GFR as the covariate. The results are summarized in terms of least square means with standard errors (SEs) and 95% confidence intervals (CIs) about treatment group differences. In cases of missing or unavailable radionuclide GFR values, GFR was estimated by cimetidine creatinine clearance. If neither was available, GFR was estimated by the MDRD formula. Substituting the estimated GFRs for the radionuclide GFRs was reasonable as the Pearson correlations were high (r = 0.95 for cimetidine creatinine clearance and r = 0.92 for the MDRD formula) and the relationships were linear. When all 3 values were not available, multiple imputation using the Predictive Model Based Method in SOLAS version 3.0 (Statistical Solutions Ltd) was used to impute GFR, with the exception that participants who commenced dialysis postrandomization were assigned a GFR value of 10 mL/min/1.73 m2. Multiple imputation was used for all other missing laboratory and vital status variables that were collected longitudinally.
Kaplan-Meier event-free survival analysis was used to plot cumulative failure curves for the composite outcome (MI, stroke, revascularization, all-cause mortality) and to estimate 36-month risks of outcome for the 2 treatment groups. Treatment effectiveness was assessed using univariate Cox proportional hazards regression model, with only treatment group in the model. Hazard ratios (HRs) with 95% CIs were calculated. A 2-tailed P < .05 was considered statistically significant.