International Variation in Vitamin Prescription and Association With Mortality in the Dialysis Outcomes and Practice Patterns Study (DOPPS) Rachel B. Fissell, MD, Jennifer L. Bragg-Gresham, MS, Brenda W. Gillespie, PhD, David A. Goodkin, MD, Juergen Bommer, MD, Akira Saito, MD, Takashi Akiba, MD, Friedrich K. Port, MD, and Eric W. Young, MD ● Background: The prevalence of water-soluble vitamin use among hemodialysis (HD) patients and whether mortality and hospitalization are associated with water-soluble vitamin use by HD patients have not previously been reported. The present study investigates patterns of water-soluble vitamin use among HD patients in the Dialysis Outcomes and Practice Patterns Study (DOPPS) I and evaluates outcomes associated with vitamin use. Methods: The study sample came from the DOPPS I, a prospective observational study of adult HD patients (N ⴝ 16,345) randomly selected from 308 representative dialysis facilities in France, Germany, Italy, Japan, Spain, the United Kingdom, and the United States. Time-dependent Cox regression models were used to assess relative risk (RR) for mortality and hospitalization for patients administered water-soluble vitamins versus those not administered water-soluble vitamins. Results: There was large variation by region in the percentage of patients administered water-soluble vitamins: Europe ranged from a low of 3.7% in the United Kingdom to a high of 37.9% in Spain; 5.6% in Japan; and 71.9% in the United States. Patient use of water-soluble vitamins was associated with a substantially and significantly lower risk for mortality (RR, 0.84; P ⴝ 0.001). Lower RR for facility-level mortality also was associated with greater water-soluble vitamin use (RR, 0.98; P ⴝ 0.05 per 10% more patients administered water-soluble vitamins at the facility). Conclusion: Although only a randomized trial could prove that water-soluble vitamins improve outcomes, use of water-soluble vitamins is a minimal-risk practice pattern associated with improved outcomes in this prospective observational study. Am J Kidney Dis 44:293-299. © 2004 by the National Kidney Foundation, Inc. INDEX WORDS: Dialysis Outcomes and Practice Patterns Study (DOPPS) I; folate; hospitalization; mortality; vitamin.

L

OWER SERUM LEVELS of water-soluble vitamins caused by removal by dialysis therapy are a concern for hemodialysis (HD) patients. Folate and the B vitamins are water soluble and are not stored in large quantities in the body. Early studies showed decreased serum levels of water-soluble vitamins in HD patients, and initial recommendations in the United States called specifically for vitamin supplementation.1,2 Currently, it is common practice in the United States to give HD patients water-soluble vitamin supplementation, usually as multivitamins that contain water-soluble vitamins.3-5 However, evidence for associations between improved patient outcomes and the use of multivitamins or specific water-soluble vitamins is scarce. Since the original recommendations in the United States for daily vitamin supplementation in HD patients, dialysis therapy has changed. High-flux high-efficiency dialysis may remove even greater amounts of water-soluble vitamins than originally reported in the early days of HD therapy.6 Recent evidence links hyperhomocysteinemia with worsened cardiac and vascular outcomes.7 The evidence is particularly relevant

to patients with chronic kidney disease and endstage renal disease (ESRD), who frequently have elevated homocysteine levels.8,9 Several recent studies suggested that vitamin supplementation with B6, B12, and folate may reduce homocysteine levels in HD patients.10-15 This recent area of investigation makes the question of outcomes

From the Division of Nephrology; and Department of Biostatistics, University of Michigan; University Renal Research and Education Association; Department of Veterans Affairs Medical Center, Ann Arbor, MI; ICOS Corp, Bothell, WA; Nephrology, University of Heidelberg, Heidelberg, Germany; Medicine, Tokai University, Kanagawa; and Medicine, Tokyo Women’s Medical University, Tokyo, Japan. Received January 28, 2004; accepted in revised form April 14, 2004. The DOPPS is supported by research grants from Amgen Inc and Kirin Brewery Ltd without restrictions on publications. R.B.F. received a research fellowship from the National Kidney Foundation. Address reprint requests to Rachel B. Fissell, MD, VA Ann Arbor Healthcare System, Nephrology 111J, Veterans Affairs Department, 2215 Fuller Rd, Ann Arbor, MI 48105-2303. E-mail: [email protected] © 2004 by the National Kidney Foundation, Inc. 0272-6386/04/4402-0028$30.00/0 doi:10.1053/j.ajkd.2004.04.047

American Journal of Kidney Diseases, Vol 44, No 2 (August), 2004: pp 293-299

293

294

FISSELL ET AL

associated with water-soluble vitamin use in HD patients highly relevant. The prevalence of water-soluble vitamin use among HD patients across countries has not been reported previously. Associations between watersoluble vitamin use and mortality and hospitalization also have not been clearly defined in the HD population. The present study investigates patterns of water-soluble vitamin use among 7 countries in the Dialysis Outcomes and Practice Patterns Study (DOPPS) I—France, Germany, Italy, Japan, Spain, the United Kingdom, and the United States—and evaluates mortality and hospitalization as associated with water-soluble vitamin use. METHODS

Data Sources This analysis used a sample of 16,345 HD patients from the DOPPS I, a prospective observational study involving adult HD patients randomly selected from 308 representative dialysis facilities. In each facility, an algorithm was used to sample from 20 to 40 patients, depending on facility size. The DOPPS I sampling plan and study methods have been described elsewhere.16 Consent to collect patient information without patient identifiers was obtained as needed from the local or national ethics committee or institutional review board. US facilities started the study in 1997; European facilities, in 1998; and Japanese facilities, in 1999. The present analysis used data gathered at 4-month follow-up intervals through spring 2001.

Classification of Vitamin Use Information about vitamin use was abstracted from patient charts by a study coordinator in each facility. Lists of all medications and supplements, both over the counter and prescription, were collected. Medications entered as multivitamins were assumed to contain water-soluble vitamins. In the United States, the team was able to evaluate whether the named multivitamins administered contained any of the water-soluble vitamins folate, B6, B12, or ascorbic acid. Use of specific fat-soluble vitamins (vitamins A, D, E, and K) was not included in this analysis.

Statistical Methods Patient demographics, comorbidities, and multivitamin use were investigated first by using descriptive statistics. To estimate the overall proportions of patient water-soluble vitamin use by country, sample weights were used to account for the differing proportions of patients sampled in each facility. This allowed our sample to represent all patients from a given facility. Logistic regression analysis was used to examine associations between multivitamin use and both patient demographics and comorbid conditions. Logistic regression analysis also was used to evaluate associations between levels of normalized protein catabolic rate

(nPCR), serum albumin concentration, serum hemoglobin concentration, neutrophil count, and water-soluble vitamin use. Generalized estimating equations were used to account for clustering at the facility level, assuming a compound symmetry covariance structure.17,18 Extra binomial facility variation was tested by using a mixed-model logistic regression, assuming a binomial distribution of multivitamin use with the same versus different probability of prescription at each facility.19 Cox proportional hazards regression models were used to examine the relationship between mortality or time to first hospitalization and multivitamin use. In addition, use of water-soluble vitamins was modeled at the HD facility level by evaluating the association between each outcome and the percentage of patients in each facility administered watersoluble vitamins. All Cox models were adjusted for age, sex, race, time on HD, and 15 summary comorbid conditions: previous transplant, positive test result for hepatitis B virus, coronary artery disease, congestive heart failure, other cardiac disease, hypertension, diabetes mellitus, cerebrovascular disease, peripheral vascular disease, cancer, human immunodeficiency virus/acquired immunodeficiency syndrome, lung disease, neurological disorder, psychiatric disease, and gastrointestinal bleeding. A time-dependent Cox model was used to allow patients to either start or stop multivitamin use during the study period. A sensitivity analysis included only patients in the sample who did not switch multivitamin status (83.3% of patients). Cox proportional hazards models were stratified by country of residence. An additional sensitivity analysis used interactions to verify that the association between mortality and patient-level water-soluble vitamin use was consistent among continents.

RESULTS

Table 1 lists percentages of the 16,345 randomly selected HD patients administered watersoluble vitamins by country. Overall patient use of water-soluble vitamins ranged from 3.7% in the United Kingdom to 71.9% in the United States. The median percentage of patients in each facility administered water-soluble vitamins was relatively low in Japan (0.0%), Italy (1.9%), and the United Kingdom (2.3%); somewhat greater in Spain (19.9%); and substantially greater in the United States (65.5%), with P ⬍ 0.0001 for variation by country. Water-soluble vitamin use was stable in each country during the years of study. Results also are given at the facility level as percentage of patients in each facility administered water-soluble vitamins. Figure 1 shows the overall distribution of water-soluble vitamin use by facility. Mean water-soluble vitamin prevalence at DOPPS facilities worldwide was 35.1%, with a median of 28.6%. The percentage of

MORTALITY AND WATER-SOLUBLE VITAMIN USE Table 1.

295

Percentage of Patients Administered Water-Soluble Vitamins, by Country Facility Use (% patients in facility)

Country*

Overall Patient Use† (%)

5th Percentile

Median

95th Percentile

France Germany Italy Japan Spain United Kingdom United States

12.5 14.5 6.4 5.6 37.9 3.7 71.9

0.0 0.0 0.0 0.0 3.7 0.0 35.7

3.4 5.9 1.9 0.0 19.9 2.3 65.5

42.2 65.7 48.7 28.2 70.7 12.4 92.8

NOTE. N ⫽ 16,345. *P ⬍ 0.0001 for variation by country. †Sampling weights used to account for facility size.

facilities that administered water-soluble vitamins to less than 10% of their patients was 42.0%, and the percentage of facilities that administered water-soluble vitamins to 70% or greater of their patients was 20.5%. Water-Soluble Vitamin Components In the United States, we were able to evaluate whether the multivitamins administered contained the water-soluble vitamins folate, B6, B12, and/or ascorbic acid. Of patients administered water-soluble vitamins in the United States, 72.2% were administered water-soluble vitamins with folate and 100% were administered watersoluble vitamins with B6, B12, and ascorbic acid. Components correlated highly with one another (correlation coefficients ⬎0.8). Therefore, it was not possible to resolve the contribution of individual components (ie, the water-soluble vitamins listed) to outcomes.

Fig 1. Overall distribution of water-soluble vitamin use by facility in the United States, Europe, and Japan. Mean water-soluble vitamin prevalence at the facility level was 35.1%, with a median of 28.6%. The percentage of facilities with a water-soluble vitamin prevalence less than 10% was 42.0%, and the percentage of facilities with a water-soluble vitamin prevalence greater than 70% was 20.5%.

Factors Associated With Water-Soluble Vitamin Use Table 2 lists demographic and comorbid characteristics of the patient sample and their associated adjusted odds of water-soluble vitamin use. Mean and median ages were 59.9 years and 62 years (not shown), respectively. The sample was 56.5% men. As in other observational studies of HD patients, such comorbid conditions as diabetes mellitus and cardiac disease were common. Black patients, compared with all other races, comprised 16.9% of the sample. The categories white, Asian, and other (not shown) were combined mainly because whites were predominantly European and American, whereas Asians were almost exclusively treated in Japan. Other race contributed only 5% of the sample. Characteristics associated with increased odds of water-soluble vitamin use included older age,

296

FISSELL ET AL

Table 2. Demographic and Comorbid Characteristics of Study Sample and Their Associated, Adjusted Odds of Water-Soluble Vitamin Use* Predictors of WaterSoluble Vitamin Use Characteristic

Mean ⫾ SD or % Sample

OR*

P

Age (per 10 y) Male (v female) Black (v all other) Time on HD (per y) Diabetes mellitus

59.9 (14.7) 56.5 16.9 4.9 (5.4) 33.1

1.04 1.08 0.90 1.05 0.87

0.004 0.068 0.065 ⬍0.0001 0.002

*Adjusted for all factors listed in this table plus country of residence, facility clustering, and 13 additional nonsignificant comorbidities (P ⬎ 0.10; see text).

longer time on HD, and, of borderline significance, male sex and nonblack race. Diabetes mellitus was associated with reduced odds of water-soluble vitamin use. Characteristics not associated with adjusted odds of water-soluble vitamin use (P ⬎ 0.1) included the comorbid conditions coronary artery disease, congestive heart failure, other forms of cardiac disease, hypertension, cerebrovascular disease, peripheral vascular disease, lung disease, cancer, human immunodeficiency virus/acquired immunodeficiency syndrome, gastrointestinal bleeding, neurological disease, psychiatric disease, and recurrent cellulitis/gangrene. Mortality and Hospitalization Table 3 lists the associations between watersoluble vitamin use and relative risk (RR) for mortality and hospitalization using patient- and facility-level analyses. Patients administered compared with those not administered water-soluble Table 3.

vitamins had a significantly lower mortality risk (RR, 0.84; P ⫽ 0.001), adjusted for age, sex, race, comorbid conditions, albumin, time on HD, average facility dialysis dose (single-pool Kt/V), body mass index, average facility nPCR, and average facility hemoglobin level. The models also account for facility clustering and are stratified by country. When only patients in the sample who did not switch water-soluble vitamin status were included (83.3% of patients), results were similar (RR, 0.88; P ⫽ 0.003; not shown). Although not significant, risk for hospitalization at the patient level also was lower among patients administered water-soluble vitamins (RR, 0.94; P ⫽ 0.24) than among those not administered water-soluble vitamins (Table 3). The association between mortality and water-soluble vitamin use did not vary by geographic region, based on interaction terms that were nonsignificant and models stratified by country. Considering the proportions of patients administered water-soluble vitamins in each facility as a predictor, associations were similar, but not as large or significant. As listed in Table 3, when grouping facilities by 10% increments of patient use of water-soluble vitamins, facility-level mortality was lower for units in which patients had greater water-soluble vitamin use (RR, 0.98; eg, for 40% versus 30% water-soluble vitamin use; P ⫽ 0.05). Hospitalization was not associated with facility-level water-soluble vitamin use (RR, 1.00; P ⫽ 0.59). Figure 2 shows a general dose-response relationship between RR for mortality and quartiles of patients in the facility prescribed water-soluble vitamins. The highest quartile had significantly lower mortality com-

Patient and Facility Associations Between Water-Soluble Vitamin Use and RR for Mortality and Hospitalization Water-Soluble Vitamin Use

Patient level (water-soluble vitamins v no water-soluble vitamins)* Mortality Hospitalization Facility level (per 10% patient use of water-soluble vitamins) Mortality Hospitalization

RR

95% Confidence Interval

P

0.84 0.94

0.76-0.94 0.85-1.04

0.001 0.24

0.98 1.00

0.95-1.00 0.99-1.01

0.05 0.59

NOTE. Adjusted for age, sex, race, comorbid conditions, albumin, time on HD, average facility dialysis dose (single-pool Kt/V), body mass index, average facility nPCR, and average facility hemoglobin level. Accounts for facility clustering and stratified by country. *Time-dependent Cox model; considers changes in multivitamin status over time in the study.

MORTALITY AND WATER-SOLUBLE VITAMIN USE

297

Fig 2. Dose-response relationship between RR for mortality by facility quartiles of water-soluble vitamin use (RR, 0.98; P ⴝ 0.05).

pared with the other 3 quartiles combined (RR, 0.90; P ⫽ 0.05; not shown). Associated Nutrition Markers Odds ratios (ORs) of increased water-soluble vitamin use were associated significantly with greater nPCR (OR, 1.04 per 0.1 g/d; P ⫽ 0.007) and albumin level of 3.5 g/d or greater (OR, 1.14; P ⫽ 0.02). The odds of increased watersoluble vitamin use also appeared to be associated with greater hemoglobin level, but not significantly so (OR, 1.09; P ⫽ 0.13). The odds of water-soluble vitamin use were significantly lower in patients with a greater serum neutrophil percentage (OR, 0.94 per 10%; P ⫽ 0.002). DISCUSSION

The present DOPPS I study shows large variation in patterns of water-soluble vitamin use among HD units within and across the 7 countries. This study also describes a strong and significant association between lower mortality and regular use of water-soluble vitamins in HD patients. There are several potential reasons for this observation. Water-soluble vitamin use, and specifically folic acid use, may decrease serum homocysteine levels and thereby decrease cardiac mortality, the leading cause of mortality in HD patients. Use of water-soluble vitamins may be accompanied by greater protein intake at the patient level, suggested by the greater nPCR among water-soluble vitamin users. Better nutrition has been shown to correlate with improved survival.20 However, the lower mortality associated with regular water-soluble vitamin use per-

sists in our models after adjusting for 2 potential markers of nutrition, serum albumin level and nPCR. It also is possible that use of watersoluble vitamins at the unit level may mark other beneficial unit-specific practice patterns, eg, more meticulous overall care at the dialysis unit. One could hypothesize that socioeconomic patient characteristics not captured by this analysis also may have a role. At least 1 ongoing prospective randomized trial is testing the hypothesis that lower mortality is associated with water-soluble vitamin use in HD patients.21 Several potential explanations for the dramatic country differences in water-soluble vitamin use exist. Differences in water-soluble vitamin cost and insurance coverage by country may explain country variation in water-soluble vitamin use. Health care providers may not regard water-soluble vitamins as medically necessary because small short-term studies of vitamin supplementation have shown no survival benefit.15,22 Patient beliefs, preferences, and customs regarding water-soluble vitamins also may vary by country and thereby contribute to the observed variation in water-soluble vitamin use. The lower mortality associated with regular water-soluble vitamin use was not clearly accompanied by lower rates of hospitalization in either the patient- or facility-level models. The patientlevel model suggests an association between lower hospitalization and regular water-soluble vitamin use, but the result was not significant. As with any survey involving medication use, this study had the potential for misclassification bias. Patients and study coordinators filling out

298

the survey forms may have carefully reported prescription drug use, but less completely reported drugs or nutritional supplements that patients purchased without a prescription, such as multivitamins. The survey was repeated in each unit every 4 months. It is likely that repeated sampling helped give a more complete record of nonprescription medication use. Furthermore, misclassification bias would lead to underestimating a beneficial effect of water-soluble vitamins. The possibility of survivor and hospitalization bias also was considered here. As described in the Results section, 83.3% of patients had no change in their water-soluble vitamin status during the study period. An analysis was run that included only patients who had no change in their water-soluble vitamin status during the study period. This analysis yielded results similar to models with the full sample, in which patients who either stopped or started water-soluble vitamin use during the study period were analyzed as changing group assignments. The associations with nutrition markers are intriguing. It is unclear whether water-soluble vitamin use could lead directly to improved nPCR, serum albumin concentration, and/or perhaps hemoglobin concentration. These positive associations may simply reflect an overall greater nutritional status of patients administered watersoluble vitamins. The nonsignificant trend for anemia possibly could be related to vitamin B12. The lower serum neutrophil percentage in patients administered water-soluble vitamins may reflect a decreased inflammatory state associated with water-soluble vitamin use. Neither a causal relationship nor the mechanism behind such a hypothesized causal relationship can be inferred from the present study. Only a randomized trial could prove that watersoluble vitamins improve outcomes. However, this prospective observational study provides international evidence that HD patient use of watersoluble vitamins is associated with reduced mortality. Water-soluble vitamins are known to be lost during HD. The practice pattern of regularly prescribing water-soluble vitamins is easily initiated and low risk, and practitioners internationally may want to consider prescribing them while awaiting more definitive data from interventional studies. Given the potentially large benefit, prospective randomized trials are needed to con-

FISSELL ET AL

firm this result, and more studies are needed to clarify the underlying reasons for the association shown. ACKNOWLEDGMENT The authors thank dialysis unit staffs and directors of the Worldwide DOPPS I facilities for contributing data and the DOPPS I country investigators for their contributions during the design and implementation of the DOPPS. For a full listing of investigators, please see reference 16.

REFERENCES 1. Sullivan JF, Eisenstein AB, Mottola OM, Mittal AK: The effect of dialysis on plasma and tissue levels of vitamin C. Trans Am Soc Artif Intern Organs 18:277-282, 1972 2. Kopple JD, Swendseid ME: Vitamin nutrition in patients undergoing maintenance hemodialysis. Kidney Int Suppl 2:S79-S84, 1975 3. Descombes E, Hanck AB, Fellay G: Water-soluble vitamins in chronic hemodialysis patients and need for supplementation. Kidney Int 43:1319-1328, 1993 4. Makoff R, Dwyer J, Rocco MV: Folic acid, pyridoxine, cobalamin, and homocysteine and their relationship to cardiovascular disease in end-stage renal disease. J Ren Nutr 6:2-11, 1996 5. Friedman AN: Pharmacologic B-vitamin therapy for hyperhomocysteinemia in dialysis patients: Has the time come? Nutr Clin Care 5:20-24, 2002 6. Kasama R, Koch T, Canals-Navas C, Pitone JM: Vitamin B6 and hemodialysis: The impact of high-flux/highefficiency dialysis and review of the literature. Am J Kidney Dis 27:680-686, 1996 7. Clarke R, Daly L, Robinson K, et al: Hyperhomocysteinemia: An independent risk factor for vascular disease. N Engl J Med 324:1149-1155, 1991 8. Moustapha A, Naso A, Nahlawi M, et al: Prospective study of hyperhomocysteinemia as an adverse cardiovascular risk factor in end-stage renal disease. Circulation 97:138141, 1998 9. Perna AF, Castaldo P, Ingrosso D, De Santo NG: Homocysteine, a new cardiovascular risk factor, is also a powerful uremic toxin. J Nephrol 12:230-240, 1999 10. Robinson K, Gupta A, Dennis V, et al: Hyperhomocysteinemia confers an independent increased risk of atherosclerosis in end-stage renal disease and is closely linked to plasma folate and pyridoxine concentrations. Circulation 94:2743-2748, 1996 11. Sunder-Plassmann G, Fodinger M, Buchmayer H, et al: Effect of high dose folic acid therapy on hyperhomocysteinemia in hemodialysis patients: Results of the Vienna multicenter study. J Am Soc Nephrol 11:1106-1116, 2000 12. Tremblay R, Bonnardeaux A, Geadah D, et al: Hyperhomocysteinemia in hemodialysis patients: Effects of the 12-month supplementation with hydrosoluble vitamins. Kidney Int 58:851-858, 2000 13. Elian KM, Hoffer LJ: Hydroxocobalamin reduces hyperhomocysteinemia in end-stage renal disease. Metabolism 51:881-886, 2002 14. Kaplan LN, Mamer OA, Hoffer LJ: Parenteral vita-

MORTALITY AND WATER-SOLUBLE VITAMIN USE

min B12 reduces hyperhomocysteinemia in end-stage disease. Clin Invest Med 24:5-11, 2001 15. Billion S, Tribout B, Cadet E, et al: Hyperhomocysteinaemia, folate and vitamin B12 in unsupplemented haemodialysis patients: Effect of oral therapy with folic acid and vitamin B12. Nephrol Dial Transplant 17:455-461, 2002 16. Young EW, Goodkin DA, Mapes DL, et al: The Dialysis Outcomes and Practice Patterns Study: An international hemodialysis study. Kidney Int 57:S74-S81, 2000 (suppl 74) 17. SAS Institute Inc: SAS/STAT User’s Guide, version 8, vol 2, Cary, NC, SAS Institute, 2000, p 1452 18. Klein J, Moeschberger M: Survival Analysis Techniques for Censored and Truncated Data, New York, NY, Springer, 1997, pp 416-418 19. SAS Technical Support: GLIMMIX: A SAS Macro

299

for Fitting Generalized Linear Mixed Models Using Proc Mixed and the Output Delivery System (ODS). Available at http://ftp.sas.com/techsup/download/stat/glmm800.html. Accessed March 24, 2004 20. Pifer TB, McCullough KP, Port FK, et al: Mortality risk in hemodialysis patients and changes in nutritional indicators: DOPPS. Kidney Int 62:2238-2245, 2002 21. Jamison R, Hartigan P, Gaziano M, et al: High plasma homocysteine (Hcy) and cardiovascular (CV) disease in end stage (ESRD) and advanced chronic (ACRD) renal disease. The VA Cooperative Study Program Homocysteine Study (HOST). J Am Soc Nephrol 13:441A, 2002 (abstr) 22. Arnadottir M, Brattstrom L, Simonsen O, et al: The effect of high-dose pyridoxine and folic acid supplementation on serum lipid and plasma homocysteine concentrations in dialysis patients. Clin Nephrol 40:236-240, 1993