- Email: [email protected]
Racial differences in survival of patients on dialysis
Kidney International, Vol. 58 (2000), pp. 1293–1299
Racial differences in survival of patients on dialysis YORK P.C. PEI, CELIA M.T. GREENWOOD, ANNE L. CHERY, and GEORGE G. WU Department of Medicine, University of Toronto, Toronto, Ontario; Department of Human Genetics, McGill University, Montreal, Quebec; and Toronto Regional Dialysis Registry, Toronto, Ontario, Canada
Racial differences in survival of patients on dialysis. Background. Recent studies have documented racial differences in the crude mortality rates of patients on dialysis. However, proper interpretation of these findings requires adjustment for potential confounders and comorbid risk factors between the racial groups. Methods. We examined the clinical data on 3752 Caucasian patients, 451 Southeast Asian patients, 322 South Asian patients, and 319 black patients who were treated with hemodialysis or peritoneal dialysis under a Universal Health Care system in Toronto and prospectively followed between 1981 and 1995. In all patients, a number of comorbid risk factors for survival was assessed at the start of dialysis and was reassessed with their outcome status (that is, continued dialysis, transplantation, death, or loss to follow-up) at least every six months. Cox proportional hazards analysis was used to fit multivariate models predicting patient survival. Pairwise comparisons of the relative hazards of death between the racial groups were performed after stratifying for cardiovascular disease, diabetes mellitus, and hypertension at the start of dialysis, and were adjusted for differences in other comorbid risk factors. Results. The risk of death in Caucasian patients was significantly increased when compared with Southeast Asian patients, South Asian patients, and black patients [multivariate relative hazards (95% CI): 1.63 (1.36 to 1.97), 1.36 (1.07 to 1.73), 1.34 (1.07 to 1.67), respectively]. Additionally, we detected an interaction between race and cigarette smoking (P ⬍ 0.004), suggesting that in the dialysis patients who smoked, whites had a higher mortality risk compared with non-whites. Conclusions. Differences in patient survival on dialysis exist between racial groups. However, the genetic and environmental determinants that underlie these differences are presently unknown.
Recent data from several end-stage renal disease (ESRD) registries have documented racial differences in the crude mortality rates of patients on dialysis [1–4]. For example, data from the United States Renal Data System indicated that between 1980 and 1994, the morKey words: cardiovascular disease, race, hemodialysis, end-stage renal disease, renal failure, peritoneal dialysis. Received for publication November 29, 1999 and in revised form April 7, 2000 Accepted for publication April 14, 2000
2000 by the International Society of Nephrology
tality rate (adjusted for age, gender, and renal disease diagnosis) in Caucasian patients was consistently higher than black patients and patients from the Asia-Pacific regions [1, 2]. Similarly, the crude mortality rate of hemodialysis patients in the age range of 45 to 54 years was 5 to 10% higher in Europe than in Japan [3, 4]. However, without adjusting for differences in dialysis accessibility and other comorbid risk factors for survival between racial groups, proper interpretation of these findings remains uncertain. Treatment of ESRD in Canada consists of a comprehensive and integrated system of government-funded programs in hemodialysis, peritoneal dialysis, and renal transplantation, which is universally accessible to all Canadians [5]. Since 1981, the Toronto Regional Dialysis Registry has been collecting data on demographics and comorbid risk factors in all patients at start of their ESRD treatment program (Methods section). This database reflects a regional experience of a population of 4.5 million, as well as a diverse mixture of ethnic groups (that is, 80% Caucasian, 10% Southeast Asian, 5% South Asian, and 5% blacks) unique to the Metropolitan Toronto area [6]. METHODS Study design and data collection This is a prospective cohort study. Since January of 1981, the Toronto Regional Dialysis Registry has documented a number of demographic and comorbid risk factors in all patients at start of their ESRD treatment program. They include age, gender, race, dialysis modality; dialysis start year; and ESRD diagnosis. Additionally, all patients were assessed for a history of cigarette smoking; hypertension; diabetes mellitus; symptomatic coronary, cerebral, or peripheral vascular disease; congestive heart failure; chronic respiratory disease; malignancy; and any serious illness that may shorten life expectancy to less than five years. The latter comorbid risk factors were reassessed with the patient outcome status (that is, continued dialysis treatment, death, transplantation, or loss to follow-up) every six months up to June of 1989 and then at monthly intervals thereafter.
1293
1294
Pei et al: Racial differences in survival of patients on dialysis
Patient classification and selection All patients were classified under five major racial groupings: (1) Caucasian (white); (2) Southeast Asian (including patients of Chinese, Japanese, Korean, and Indo-Chinese origins); (3) South Asian (including patients of East Indian, Pakistani, and Punjabi origins); (4) black (including black patients from Africa and the Caribbean); and (5) others (including patients of Aboriginal, Arabic, Hispanic, Polynesian, mixed, or unknown ethnic origins). All patients assessed between January 1981 and June 1995 were included in the analysis, except those who died within the first three months of dialysis, those who received a renal allograft without ever being dialyzed, and those who belonged to the “other” racial group (that is, ⬍4% of all patients). Statistical analysis All data are expressed as mean ⫾ SD or percentages and were compared between the four racial groups using one-way analysis of variance (ANOVA) for continuous variables and contingency table analysis for binary variables. Only one overall P value is given for these descriptive data to indicate whether there are differences among the racial groups. Patient survival for the four racial groups was estimated as a function of time by the Kaplan–Meier method, with patient follow-up censored at renal transplantation [7]. The Cox proportional hazards analysis was used to fit multivariate models predicting survival and to obtain multivariate relative hazards of death between the racial groups [8]. All demographic and comorbid risk factors collected as well as a composite index of cardiovascular disease severity (NRISK: 0 to 4 with the presence of angina, myocardial infarction, symptomatic peripheral, or cerebral vascular disease at start of dialysis each scored one point) were tested for a linear effect on survival and used in the analysis (Methods section). Smoothed plots of martingale residuals showed that linear covariate effects for age, start year, and NRISK adequately fit the data [9]. The multivariate model fitting included three stages. First, to control for confounding caused by other risk factors, all available covariates and two-factor interactions, excluding race, were examined, but with patient follow-up censored at transplantation at this step. Because of a lack of proportionality over time for the hazard of death between hypertensive and nonhypertensive patients and to allow for maximum flexibility in the baseline hazard, multivariate models were stratified for three known risk factors: hypertension, cardiovascular disease, and diabetes mellitus (all at the start of dialysis). All main effects except race were forced into the model, and a forward selection approach was used to examine two-way interactions between risk factors. Second, having chosen the best model predicting
survival without using any race information, race indicators were added to the model, and interactions between race and the risk factors were examined by using stepwise selection. The starting model included the main effects (including race), all the two-way risk factor interactions identified in the forward selection, and all interactions between white race and the other risk factors. The main effects, such as age, gender, and previously identified interactions, were not allowed to leave the model. Third, the effect of renal transplantation on these relative hazards was also investigated. A model with a time-dependent covariate for transplant status investigated interactions between race and transplant. Also, a multivariate model predicting time to death post-transplantation was fit. There were many fewer deaths post-transplantation so that model fitting for these data was simpler. This model was stratified by cardiovascular disease and diabetes mellitus only, because of a very small number of deaths in some strata when hypertension was included as a third stratification variable. Only a few covariates were included as predictors. We also modeled the total follow-up time, including a time-dependent variable for transplantation status, and examined interactions between this indicator, race, and other variables. To examine the relationships between smoking and underlying cardiovascular disease to the overall risk of deaths further, models were also fit to two subgroups: patients with and without cardiovascular disease at the start of dialysis. The same modeling approach as that previously mentioned was applied, first examining all predictors but race, then including the race indicators and interactions, and finally exploring the effects of transplantation. RESULTS Patient characteristics by racial groups Overall, 30% of the patient cohort remained alive on dialysis; 21% were alive with a renal allograft; 43.5% died while on dialysis or after renal transplantation; and 5.5% were lost to follow-up at the end of the study period. A total of 26% of the patient cohort received a renal allograft during the study period. Tables 1 and 2 show that a number of demographic and comorbid risk factors were different between the racial groups at start of dialysis. For example, both Southeast Asian and white patients were older than South Asian and black patients. There were more males in white and South Asian patients than Southeast Asian and black patients. More Southeast Asian and South Asian patients were on peritoneal dialysis treatment compared with whites and blacks. The prevalence of diabetes mellitus was higher in Southeast Asian, South Asian, and black patients than white patients, reflecting an increased prevalence of type II diabetes mellitus in the former racial groups. Of note,
1295
Pei et al: Racial differences in survival of patients on dialysis Table 1. Patient characteristics at start of dialysis
Age years Age ⬎65 years Sex male:female Modality PD:HDa Dialysis duration years Total follow-upb years History of smokingc Coronary artery diseased Vascular diseasee Congestive heart failuref Hypertension Malignancy Serious illnessg
Caucasian (N ⫽ 3752)
Southeast Asian (N ⫽ 451)
South Asian (N ⫽ 322)
Blacks (N ⫽ 319)
P value
56.5 ⫾ 16 36% 1.57 1.80 2.03 ⫾ 1.93 3.45 ⫾ 3.32 36% 24% 21% 26% 73% 7.5% 3.6%
57.3 ⫾ 17 42% 1.04 2.47 2.25 ⫾ 1.85 3.36 ⫾ 3.06 16% 14% 12% 24% 77% 3% 5.1%
52.7 ⫾ 15 22% 1.46 2.22 1.91 ⫾ 1.56 3.03 ⫾ 2.84 13% 23% 13% 28% 81% 3% 3.4%
50.6 ⫾ 15 18% 1.26 1.55 2.33 ⫾ 2.02 3.34 ⫾ 3.07 23% 17% 19% 27% 84% 3% 5.0%
⬍0.0001 ⬍0.0001 ⬍0.0003 ⬍0.005 ⫽0.004 ⫽0.16 ⬍0.0001 ⬍0.0001 ⬍0.0001 ⫽0.56 ⬍0.0001 ⬍0.0001 ⫽0.26
a
PD, peritoneal dialysis; HD, hemodialysis Total follow-up time since initiation of dialysis, including follow-up time after transplantation Ex- or current smoker d Coronary artery disease (history of angina and/or myocardial infarction) e Peripheral vascular disease (claudication or ischemic gangrene) or cerebral vascular disease (i.e., transient ischemic attack or stroke) f Including pulmonary edema within 6 months of starting dialysis g Serious illness limiting life expectancy to less than 5 years b c
Table 2. Primary diagnosis of end-stage renal disease
Diabetes mellitus Glomerulonephritis Hypertension Polycystic kidney disease Miscellaneous
Caucasian (N ⫽ 3752)
Southeast Asian (N ⫽ 451)
South Asian (N ⫽ 322)
Blacks (N ⫽ 319)
27% 21% 11% 7% 34%
32% 32% 11% 1% 24%
38% 25% 11% 3% 23%
37% 23% 15% 5% 20%
there was a higher prevalence of cardiovascular disease, cigarette smokers (both ex-and current smokers), and malignancies in the white patients at start of dialysis compared with the other racial groups. Table 3 shows the major causes of deaths in the four racial groups, with and without censoring at renal transplantation. Overall, 2104 patients died during the study period, and 90.6% of these deaths occurred during dialysis. In general, cardiovascular disease complications account for about half of the deaths in all racial groups. With the exception of a lower percentage of death from withdrawal of treatment in the black patients, there was no significant difference in the percentages of death in each of the main causes of death between the racial groups. Figure 1 shows the Kaplan–Meier survival curves of the four racial groups, showing all deaths that occurred during dialysis. Caucasian patients appeared to have the lowest survival among the four racial groups. The overall crude mortality rate during dialysis in white patients compared with Southeast Asian, South Asian, and black patients was 211 versus 128, 120, and 118 deaths per 1000 patientyears on dialysis, respectively. Multivariate models predicting patient survival In the best fitting models, all main effects except for gender had substantial power to predict patient survival.
The models were stratified by cardiovascular disease, hypertension, and diabetes mellitus, which all had large effects on survival. For parameters estimated in the proportional hazards regression, age, dialysis start year, malignancy, and congestive heart failure had the largest effects (in models with or without interactions), all with P values much less than 0.0001. In addition, most of the predictor variables were involved in multiple two-way interactions, suggesting that these interactions identify a high-risk patient profile that no one variable can independently capture. Having chosen the best model predicting survival without using any race information, race indicators were added to the model, and interactions between race and the risk factors were identified by using stepwise selection. The final models included the main effects (including race), all of the two-way risk factor interactions identified in the forward selection, and all interactions between white race and the other risk factors. The main effects, such as age, gender, and previously identified interactions, were not allowed to leave the model even if they were no longer significant at P ⬍ 0.05. All twoway interactions that entered each reported model are listed in Tables 4 and 5. In a few cases, the P values for the two-way interactions are greater than 0.05. This is
1296
Pei et al: Racial differences in survival of patients on dialysis Table 3. Causes of death of study patients
Deaths during dialysisa Cardiovascularb Sepsis Withdrawal from treatment Malignancy Miscellaneous All deathsc Cardiovascularb Sepsis Withdrawal from treatment Malignancy Miscellaneous a b c
Caucasian (N ⫽ 1614)
Southeast Asian (N ⫽ 130)
South Asian (N ⫽ 74)
Blacks (N ⫽ 88)
49% 15% 13% 7% 16%
48% 15% 15% 6% 16%
53% 14% 10% 5% 18%
56% 15% 6% 7% 16%
Caucasian (N ⫽ 1787)
Southeast Asian (N ⫽ 142)
South Asian (N ⫽ 83)
Blacks (N ⫽ 92)
48% 15% 13% 7% 17%
47% 14% 13% 7% 19%
51% 15% 10% 5% 19%
57% 14% 5% 8% 16%
Excluding all deaths after renal transplantation in patients who received a renal allograft Myocardial infarction, peripheral or cerebral vascular disease Including all deaths during dialysis and after renal transplantation in patients who received a renal allograft
Fig. 1. Kaplan–Meier survival curves of patients from the four racial groups, with patient follow-up censored at renal transplantation. Symbols are: (———) white; (··········) South Asian; (– – – –) South East Asian; (- - - - -) black. Caucasian patients had a lower overall survival rate compared with the other three racial groups (P ⬍ 0.0001).
because the best model, which did not include the race indicator, was founded before adding race and interactions with other variables. Of interest, a significant interaction between race and cigarette smoking was also detected (P ⬍ 0.004; see also the next section). Relative hazards of death between racial groups From the multivariate analyses, we derived pairwise relative hazards of death between the four racial groups. There was no evidence that the relative hazard of death associated with race changed with time. We found that white patients had an increased hazard of death during dialysis when compared with the three other racial groups (Table 4). Since a significant interaction between race and cigarette smoking had been detected, we also examined the relative hazards of death by smoking sta-
tus. We found that white smokers had a significantly increased hazard of death compared with smokers from the three other racial groups. For nonsmokers, white patients had a smaller but also increased hazard of death compared with patients from Southeast Asia and a trend for increased hazards of death compared with the South Asian and black patients (Table 4). Because the sample size in the non-white racial groups was relatively small, we were unable to determine whether there was any difference in the relative hazards of death between these racial groups (data not shown). We also addressed a potential concern that our findings might have been attributed to a bias in which a higher proportion of healthy Caucasians relative to the other racial groups was selected for renal transplantation. Such a bias could have led to a larger pool of
1297
Pei et al: Racial differences in survival of patients on dialysis Table 4. Relative hazards of death in study patients by racial groups Overall Death censored at renal transplantationa Caucasian vs. Southeast Asian Caucasian vs. South Asian Caucasian vs. black Caucasian vs. all Death post-transplantationa,b Caucasian vs. Southeast Asian Caucasian vs. South Asian Caucasian vs. black Caucasian vs. all
Nonsmoker
1.63 1.36 1.34 1.47
(1.36–1.97) (1.07–1.73) (1.07–1.67) (1.29–1.68)
1.12 1.04 2.44 1.30
(0.62–2.04) (0.53–2.04) (0.90–6.61) (0.85–1.99)
1.51 1.24 1.16 1.34
(1.24–1.83) (0.97–1.58) (0.92–1.47) (1.16–1.55)
Smoker 2.64 2.16 2.04 2.26
(1.75–3.98) (1.45–3.23) (1.39–2.97) (1.60–3.18)
a Relative hazards (95% CI) from Cox proportional hazards analysis stratified for cardiovascular disease, diabetes mellitus, and hypertension at start of dialysis and adjusted for all the main effects (age, gender, dialysis modality, dialysis start-year, serious illness at start of dialysis, cigarette smoking, malignancy, congestive heart failure, chronic obstructive lung disease, NRISK) and the following two-way interactions, all with P ⬍ 0.05 (hypertension ⫻ smoking; age ⫻ diabetes mellitus; diabetes mellitus ⫻ smoking; cardiovascular disease ⫻ malignancy; age ⫻ chronic obstructive lung disease; age ⫻ serious illness at start of dialysis; age ⫻ malignancy; age ⫻ NRISK; dialysis modality ⫻ serious illness at start of dialysis; dialysis start year ⫻ smoking; serious illness at start of dialysis ⫻ NRISK). b Deaths post-transplantation only in all patients who received a renal allograft. Analysis is stratified for cardiovascular disease, diabetes mellitus, and hypertension at start of dialysis, and adjusted for age, gender, transplant year, and malignancy.
Table 5. Relative hazardsa of death in study patients with or without clinical cardiovascular disease at initiation of dialysis Death censored at renal transplantation c Presented with cardiovascular disease
b
Presented without cardiovascular disease Caucasian Caucasian Caucasian Caucasian
vs. vs. vs. vs.
Southeast Asian South Asian black all
1.79 1.33 1.13 1.50
(0.43–2.26) (0.94–1.88) (0.85–1.51) (1.26–1.79)
Nonsmoker 1.20 1.27 1.33 1.26
(0.86–1.68) (0.90–1.80) (0.91–1.93) (1.00–1.59)
Smoker 2.39 2.53 2.64 2.53
(1.42–4.02) (1.49–4.28) (1.63–4.26) (1.62–3.92)
a Relative hazards (95% CI) from Cox proportional hazards analysis stratified for diabetes mellitus and hypertension at start of dialysis, and adjusted for all covariates that predict survival b For patients without cardiovascular disease at start of dialysis, the analyses were also adjusted for all the main effects as above, except NRISK and the following two-way interactions (age ⫻ diabetes mellitus, P ⫽ 0.00001; age ⫻ malignancy, P ⬍ 0.00001; hypertension ⫻ smoking, P ⬍ 0.005; age ⫻ serious illness at start of dialysis, P ⬍ 0.006; diabetes mellitus ⫻ malignancy, P ⬍ 0.01; dialysis modality ⫻ chronic obstructive lung disease, P ⬍ 0.05; hypertension ⫻ gender, P ⬍ 0.05; age ⫻ chronic obstructive lung disease, P ⬍ 0.05; diabetes mellitus ⫻ dialysis start-year, P ⬍ 0.01; dialysis start-year ⫻ smoking, P ⬍ 0.05; hypertension ⫻ chronic obstructive lung disease, P ⬍ 0.05) c For patients with cardiovascular disease at start of dialysis, the analyses were also adjusted for all of the main effects (age, gender, dialysis modality, dialysis start-year, serious illness at start of dialysis, cigarette smoking, malignancy, congestive heart failure, chronic obstructive lung disease, NRISK) and the following twoway interactions (dialysis modality ⫻ serious illness at start of dialysis, P ⬍ 0.00001; diabetes mellitus ⫻ chronic obstructive lung disease, P ⬍ 0.025; chronic obstructive lung disease ⫻ congestive heart failure, P ⬍ 0.02; hypertension ⫻ NRISK, P ⫽ 0.078).
sicker white patients on dialysis, thus accounting for the increased relative hazard of death among whites. A model that included a time-dependent covariate for transplant status did not detect a significant change in hazard with transplant for any racial group. Also, the relative hazards of death post-transplantation in our white patients were at or above unity compared with the other racial groups (Table 4). Taken together with the finding that only 9.4% of all deaths occurred after transplantation (Results section), these data provide strong evidence against a race-specific transplantation bias as an explanation for the increased relative hazards of death in our white patients. To examine the relationship between cigarette smoking and underlying cardiovascular disease to the overall risk of death further, a subgroup analysis was performed to compare the hazards of death between the four racial groups, with and without symptomatic cardiovascular
disease at the start of dialysis. Table 5 shows that in patients with symptomatic cardiovascular disease, white smokers had a significantly increased hazard of death compared with smokers from the other three racial groups. In contrast, for nonsmokers with or without symptomatic cardiovascular disease, there was a smaller and often nonsignificant trend for increased hazards of death in the white patients compared with the other racial groups. However, the overall relative hazard of death for Caucasians versus all other racial groups did not differ from the same racial comparison among patients who presented without cardiovascular disease (that is, 1.47 vs. 1.50, respectively; Tables 4 and 5). DISCUSSION Using a database that reflects the regional experience of a universal healthcare treatment system for ESRD
1298
Pei et al: Racial differences in survival of patients on dialysis
and a unique mixture of patients with different racial origins [5, 6], we have documented an increased mortality risk during dialysis in white patients compared with three other racial groups, after adjusting for differences in their comorbid risk factors. We have also excluded a potential concern that our findings might have been attributed to a bias in which a higher proportion of healthy whites were selected for transplantation, thus leaving a relatively larger pool of sicker white patients on dialysis. Atherosclerosis is a frequent and important cause of morbidity and mortality for patients with ESRD [10–16]. The notion that atherosclerosis is accelerated in the uremic milieu has been suggested [12–14] and is supported by the documented 15- to 20-fold increase in cardiovascular mortality in dialysis patients compared with ageand sex-matched populations without renal disease [10, 11, 15]. Thus, it is not surprising that cardiovascular disease accounts for 50% of all deaths in the patients with ESRD [11, 15, 16]. However, while a higher prevalence of cardiovascular disease at the start of dialysis was noted in whites compared with other racial groups (Table 1), this did not explain the observed differences in survival. One limitation of the current study was that only historic documentation of comorbid factors was used, and quantitative measures for the severity of these factors were not available. This may potentially limit our ability to detect whether the observed mortality differences could be related to different cardiovascular disease burden between the racial groups. Future studies that incorporate quantitative measures of cardiovascular disease severity are needed to clarify this issue further. Caucasian smokers with symptomatic cardiovascular disease at the start of dialysis constitute a very highrisk group for increased mortality. The basis for this observation is also presently unknown. Another limitation of the current study is that quantitative data on cigarette smoking in the study patients were not available. Thus, we cannot exclude the possibility that the mortality differences might have been related to racial differences in the cumulative dose of cigarette smoking, with the whites being the heaviest smokers. However, in a recent study of 144 healthy subjects without hypertension, hyperlipidemia, and diabetes mellitus, white smokers were found to have an increased impairment of arterial endothelial function compared with their Chinese counterparts when the dose of cigarette smoked was controlled [17]. These data suggest the possibility for a race-specific differential susceptibility to cardiovascular disease complications that may be related to environmental and/or genetic factors [17–21]. For example, racial differences in environmental factors such as dietary fat and antioxidant intake could have an important impact on cardiovascular disease mortality between the racial groups. However, comprehensive studies addressing the role of environmental factors on multiracial
cardiovascular disease burden are not presently available. Similarly, racial differences in genetic susceptibility factors may underlie the biologic variability of human disease traits such as cardiovascular disease complications [18]. In this context, a recent study has documented pharmacogenetic differences in nicotine metabolism between black and white smokers and suggests a differential risk for addiction and other medical complications of cigarette smoking between these racial groups [20, 21]. In the current study, nonsmoking white patients also had an increased hazard of death compared with Southeast Asian patients and a trend for increased hazards of death compared with South Asian and black patients (Table 4). Moreover, the overall hazard of death for whites versus all other racial groups did not differ from the same comparison among patients without symptomatic cardiovascular disease at the start of dialysis (that is, 1.47 vs. 1.5, respectively). These findings suggest that additional risk factors other than cigarette smoking and cardiovascular disease may influence survival. Of interest, dialysis dose has been suggested to be an important determinant for both morbidity and mortality in uremic patients [22, 23]. Furthermore, a reduced dialysis dose has been shown to be associated with increased mortality across all causes of death [24]. Since Southeast Asian and South Asian patients have a smaller body size and receive a higher average dialysis dose compared with Caucasian and black patients (abstract; Pei et al, J Am Soc Nephrol 8:208, 1997), racial differences in the dialysis dose could be an important determinant for the observed findings. On the other hand, a large registry study has failed to document that dialysis dose influences patient survival [25]. In addition to dialysis dose, the influence of race-specific differences in diet and lifestyle may be also important. All of these factors remain to be explored in future studies. In conclusion, we have documented an increased mortality risk in Caucasian patients with ESRD compared with three other racial groups, after adjusting for differences in their comorbid risk factors. At this time, we are unable to determine whether these mortality differences may be in part related to differences in the cardiovascular disease burden between these racial groups. However, white smokers with symptomatic cardiovascular disease at the start of dialysis constitute a very high-risk group for increased mortality compared with smokers from other racial groups. While our study has focused on uremic patients, racial differences in survival have been documented in the general population as well [26]. It is likely that both genetic and environmental factors may underlie these race-specific differences, but very little is currently known about these factors. Future studies to elucidate these race-specific, gene-environmental determinants of health and disease are greatly needed.
Pei et al: Racial differences in survival of patients on dialysis
ACKNOWLEDGMENTS We thank Dr. Stanley Fenton for his support of the Toronto Regional Dialysis Registry.
13.
Reprint requests to York Pei, M.D., Division of Nephrology, University Health Network, 13 EN-228, 200 Elizabeth Street, Toronto, Ontario, Canada, M5G 2C4. E-mail: [email protected]
14.
15.
REFERENCES 1. Agodoa L, Eggers P: Renal replacement therapy in the United States: Data from the United States Renal Data System. Am J Kidney Dis 25:119–133, 1995 2. USRDS 1997 Annual Data Report: Patient mortality and survival. Am J Kidney Dis 30:S91–S96, 1997 3. Mallick N, Jones E, Selwood N: The European (European Dialysis and Transplantation Association-European Renal Association) Registry. Am J Kidney Dis 25:176–187, 1995 4. Teraoka S, Toma H, Nihei H, Ota K, Babazono T, Ishikawa I, Shinoda A, Maeda K, Koshikawa S, Takahashi T, Sonoda T: Current status of renal replacement therapy in Japan. Am J Kidney Dis 25:151–164, 1995 5. Fenton S, Desmeules M, Copleston P, Arbus G, Froment D, Jeffrey J, Kjellstrand C: Renal replacement therapy in Canada: A report from the Canadian Organ Replacement Registry. Am J Kidney Dis 25:134–150, 1995 6. Statistics Canada: Ethnic origin, in 1991 Census Canada Ottawa: Industry, Science and Technology Canada (Catalogue Number 93– 315), 1993, pp 44–59 7. Kaplan E, Meier P: Nonparametric estimation from incomplete observations. J Am Stat Assoc 53:457–481, 1958 8. Armitage P, Berry G: Survival analysis, in Statistical Methods in Medical Research (2nd ed), Oxford, Blackwell Scientific Publications, 1987, pp 421–439 9. Therneau TM: Extending the Cox model, in Technical Report Number 58, pp 1–64, 1996, Rochester, Section of Biostatistics, Mayo Clinic 10. Wheeler D: Cardiovascular disease in patients with chronic renal failure. Lancet 348:1673–1674, 1996 11. London G, Drueke T: Atherosclerosis and arteriosclerosis in chronic renal failure. Kidney Int 51:1678–1695, 1997 12. Lindner A, Charra B, Sherrard D, Scribner B: Accelerated
16. 17.
18. 19.
20. 21. 22. 23.
24.
25. 26.
1299
atherosclerosis in prolonged maintenance hemodialysis. N Engl J Med 290:697–701, 1974 Ibels L, Alfrey A, Hufer W, Craswell P, Anderson J, Weil R: Arterial calcification and pathology in uremic patients undergoing dialysis. Am J Med 66:790–796, 1979 Kawagishi T, Nishizawa Y, Konishi T, Kawasaki K, Emoto M, Shoji T, Tabata T, Inoue T, Morii H: High-resolution ultrasonography in evaluation of atherosclerosis in uremia. Kidney Int 48:820– 826, 1995 Raine A, Margreiter R, Brunner F, Ehrich J, Geerlings W, Landais P, Mallick N, Selwood N, Tufveson G, Valderrabano F. Report on management of renal failure in Europe, XXII, 1991. Nephrol Dial Transplant 2:15–23, 1992 Parfrey P, Foley R, Harnett J, Kent G, Murray D, Barre P: Outcome and risk factors of ischemic heart disease in chronic uremia. Kidney Int 49:1428–1434, 1996 Woo K, Robinson J, Chook P, Adams M, Yip G, Mai Z, Lam C, Sorensen K, Deanfield J, Celermajer D: Differences in the effect of cigarette smoking on endothelial function in Chinese and white adults. Ann Intern Med 127:372–375, 1997 Schriver M: Ethnic variation as a key to the biology of human disease. Ann Intern Med 127:401–403, 1997 Wang X, Sim A, Badenhop, McCredie M, Wilken D: A smokingdependent risk of coronary artery disease associated with a polymorphism of the endothelial nitric oxide synthase gene. Nat Med 2:41–45, 1996 Sellers E: Pharmacogenetics and ethnoracial differences in smoking. JAMA 280:179–180, 1998 Perez-Stable E, Herrera B, Jacob P III, Benowitz N: Nicotine metabolism and intake in black and white smokers. JAMA 280: 152–156, 1988 Lowrie E, Laird N, Parker T, Sargent J: Effect of the hemodialysis prescription on patient morbidity. N Engl J Med 305:1176–1181, 1981 Owen W, Lew N, Liu Y, Lowrie E, Lazarus M: The urea reduction ratio and serum albumin concentration as predictors of mortality in patients undergoing hemodialysis. N Engl J Med 329:1001– 1006, 1993 Bloembergen W, Stannard D, Port F, Wolfe R, Pugh J, Jones C, Greer JW, Golper TA, Held PJ: Relationship of dose of hemodialysis and cause-specific mortality. Kidney Int 50:557–565, 1996 Owen W, Chertow G, Lazarus J, Lowrie E: Dose of hemodialysis and survival: Differences by race and sex. JAMA 280:1764–1768, 1998 Hahn R, Eberhardt S: Life expectancy in four U.S. racial/ethnic population. Epidemiology 6:350–355, 1995
Racial differences in survival of patients on dialysis YORK P.C. PEI, CELIA M.T. GREENWOOD, ANNE L. CHERY, and GEORGE G. WU Department of Medicine, University of Toronto, Toronto, Ontario; Department of Human Genetics, McGill University, Montreal, Quebec; and Toronto Regional Dialysis Registry, Toronto, Ontario, Canada
Racial differences in survival of patients on dialysis. Background. Recent studies have documented racial differences in the crude mortality rates of patients on dialysis. However, proper interpretation of these findings requires adjustment for potential confounders and comorbid risk factors between the racial groups. Methods. We examined the clinical data on 3752 Caucasian patients, 451 Southeast Asian patients, 322 South Asian patients, and 319 black patients who were treated with hemodialysis or peritoneal dialysis under a Universal Health Care system in Toronto and prospectively followed between 1981 and 1995. In all patients, a number of comorbid risk factors for survival was assessed at the start of dialysis and was reassessed with their outcome status (that is, continued dialysis, transplantation, death, or loss to follow-up) at least every six months. Cox proportional hazards analysis was used to fit multivariate models predicting patient survival. Pairwise comparisons of the relative hazards of death between the racial groups were performed after stratifying for cardiovascular disease, diabetes mellitus, and hypertension at the start of dialysis, and were adjusted for differences in other comorbid risk factors. Results. The risk of death in Caucasian patients was significantly increased when compared with Southeast Asian patients, South Asian patients, and black patients [multivariate relative hazards (95% CI): 1.63 (1.36 to 1.97), 1.36 (1.07 to 1.73), 1.34 (1.07 to 1.67), respectively]. Additionally, we detected an interaction between race and cigarette smoking (P ⬍ 0.004), suggesting that in the dialysis patients who smoked, whites had a higher mortality risk compared with non-whites. Conclusions. Differences in patient survival on dialysis exist between racial groups. However, the genetic and environmental determinants that underlie these differences are presently unknown.
Recent data from several end-stage renal disease (ESRD) registries have documented racial differences in the crude mortality rates of patients on dialysis [1–4]. For example, data from the United States Renal Data System indicated that between 1980 and 1994, the morKey words: cardiovascular disease, race, hemodialysis, end-stage renal disease, renal failure, peritoneal dialysis. Received for publication November 29, 1999 and in revised form April 7, 2000 Accepted for publication April 14, 2000
2000 by the International Society of Nephrology
tality rate (adjusted for age, gender, and renal disease diagnosis) in Caucasian patients was consistently higher than black patients and patients from the Asia-Pacific regions [1, 2]. Similarly, the crude mortality rate of hemodialysis patients in the age range of 45 to 54 years was 5 to 10% higher in Europe than in Japan [3, 4]. However, without adjusting for differences in dialysis accessibility and other comorbid risk factors for survival between racial groups, proper interpretation of these findings remains uncertain. Treatment of ESRD in Canada consists of a comprehensive and integrated system of government-funded programs in hemodialysis, peritoneal dialysis, and renal transplantation, which is universally accessible to all Canadians [5]. Since 1981, the Toronto Regional Dialysis Registry has been collecting data on demographics and comorbid risk factors in all patients at start of their ESRD treatment program (Methods section). This database reflects a regional experience of a population of 4.5 million, as well as a diverse mixture of ethnic groups (that is, 80% Caucasian, 10% Southeast Asian, 5% South Asian, and 5% blacks) unique to the Metropolitan Toronto area [6]. METHODS Study design and data collection This is a prospective cohort study. Since January of 1981, the Toronto Regional Dialysis Registry has documented a number of demographic and comorbid risk factors in all patients at start of their ESRD treatment program. They include age, gender, race, dialysis modality; dialysis start year; and ESRD diagnosis. Additionally, all patients were assessed for a history of cigarette smoking; hypertension; diabetes mellitus; symptomatic coronary, cerebral, or peripheral vascular disease; congestive heart failure; chronic respiratory disease; malignancy; and any serious illness that may shorten life expectancy to less than five years. The latter comorbid risk factors were reassessed with the patient outcome status (that is, continued dialysis treatment, death, transplantation, or loss to follow-up) every six months up to June of 1989 and then at monthly intervals thereafter.
1293
1294
Pei et al: Racial differences in survival of patients on dialysis
Patient classification and selection All patients were classified under five major racial groupings: (1) Caucasian (white); (2) Southeast Asian (including patients of Chinese, Japanese, Korean, and Indo-Chinese origins); (3) South Asian (including patients of East Indian, Pakistani, and Punjabi origins); (4) black (including black patients from Africa and the Caribbean); and (5) others (including patients of Aboriginal, Arabic, Hispanic, Polynesian, mixed, or unknown ethnic origins). All patients assessed between January 1981 and June 1995 were included in the analysis, except those who died within the first three months of dialysis, those who received a renal allograft without ever being dialyzed, and those who belonged to the “other” racial group (that is, ⬍4% of all patients). Statistical analysis All data are expressed as mean ⫾ SD or percentages and were compared between the four racial groups using one-way analysis of variance (ANOVA) for continuous variables and contingency table analysis for binary variables. Only one overall P value is given for these descriptive data to indicate whether there are differences among the racial groups. Patient survival for the four racial groups was estimated as a function of time by the Kaplan–Meier method, with patient follow-up censored at renal transplantation [7]. The Cox proportional hazards analysis was used to fit multivariate models predicting survival and to obtain multivariate relative hazards of death between the racial groups [8]. All demographic and comorbid risk factors collected as well as a composite index of cardiovascular disease severity (NRISK: 0 to 4 with the presence of angina, myocardial infarction, symptomatic peripheral, or cerebral vascular disease at start of dialysis each scored one point) were tested for a linear effect on survival and used in the analysis (Methods section). Smoothed plots of martingale residuals showed that linear covariate effects for age, start year, and NRISK adequately fit the data [9]. The multivariate model fitting included three stages. First, to control for confounding caused by other risk factors, all available covariates and two-factor interactions, excluding race, were examined, but with patient follow-up censored at transplantation at this step. Because of a lack of proportionality over time for the hazard of death between hypertensive and nonhypertensive patients and to allow for maximum flexibility in the baseline hazard, multivariate models were stratified for three known risk factors: hypertension, cardiovascular disease, and diabetes mellitus (all at the start of dialysis). All main effects except race were forced into the model, and a forward selection approach was used to examine two-way interactions between risk factors. Second, having chosen the best model predicting
survival without using any race information, race indicators were added to the model, and interactions between race and the risk factors were examined by using stepwise selection. The starting model included the main effects (including race), all the two-way risk factor interactions identified in the forward selection, and all interactions between white race and the other risk factors. The main effects, such as age, gender, and previously identified interactions, were not allowed to leave the model. Third, the effect of renal transplantation on these relative hazards was also investigated. A model with a time-dependent covariate for transplant status investigated interactions between race and transplant. Also, a multivariate model predicting time to death post-transplantation was fit. There were many fewer deaths post-transplantation so that model fitting for these data was simpler. This model was stratified by cardiovascular disease and diabetes mellitus only, because of a very small number of deaths in some strata when hypertension was included as a third stratification variable. Only a few covariates were included as predictors. We also modeled the total follow-up time, including a time-dependent variable for transplantation status, and examined interactions between this indicator, race, and other variables. To examine the relationships between smoking and underlying cardiovascular disease to the overall risk of deaths further, models were also fit to two subgroups: patients with and without cardiovascular disease at the start of dialysis. The same modeling approach as that previously mentioned was applied, first examining all predictors but race, then including the race indicators and interactions, and finally exploring the effects of transplantation. RESULTS Patient characteristics by racial groups Overall, 30% of the patient cohort remained alive on dialysis; 21% were alive with a renal allograft; 43.5% died while on dialysis or after renal transplantation; and 5.5% were lost to follow-up at the end of the study period. A total of 26% of the patient cohort received a renal allograft during the study period. Tables 1 and 2 show that a number of demographic and comorbid risk factors were different between the racial groups at start of dialysis. For example, both Southeast Asian and white patients were older than South Asian and black patients. There were more males in white and South Asian patients than Southeast Asian and black patients. More Southeast Asian and South Asian patients were on peritoneal dialysis treatment compared with whites and blacks. The prevalence of diabetes mellitus was higher in Southeast Asian, South Asian, and black patients than white patients, reflecting an increased prevalence of type II diabetes mellitus in the former racial groups. Of note,
1295
Pei et al: Racial differences in survival of patients on dialysis Table 1. Patient characteristics at start of dialysis
Age years Age ⬎65 years Sex male:female Modality PD:HDa Dialysis duration years Total follow-upb years History of smokingc Coronary artery diseased Vascular diseasee Congestive heart failuref Hypertension Malignancy Serious illnessg
Caucasian (N ⫽ 3752)
Southeast Asian (N ⫽ 451)
South Asian (N ⫽ 322)
Blacks (N ⫽ 319)
P value
56.5 ⫾ 16 36% 1.57 1.80 2.03 ⫾ 1.93 3.45 ⫾ 3.32 36% 24% 21% 26% 73% 7.5% 3.6%
57.3 ⫾ 17 42% 1.04 2.47 2.25 ⫾ 1.85 3.36 ⫾ 3.06 16% 14% 12% 24% 77% 3% 5.1%
52.7 ⫾ 15 22% 1.46 2.22 1.91 ⫾ 1.56 3.03 ⫾ 2.84 13% 23% 13% 28% 81% 3% 3.4%
50.6 ⫾ 15 18% 1.26 1.55 2.33 ⫾ 2.02 3.34 ⫾ 3.07 23% 17% 19% 27% 84% 3% 5.0%
⬍0.0001 ⬍0.0001 ⬍0.0003 ⬍0.005 ⫽0.004 ⫽0.16 ⬍0.0001 ⬍0.0001 ⬍0.0001 ⫽0.56 ⬍0.0001 ⬍0.0001 ⫽0.26
a
PD, peritoneal dialysis; HD, hemodialysis Total follow-up time since initiation of dialysis, including follow-up time after transplantation Ex- or current smoker d Coronary artery disease (history of angina and/or myocardial infarction) e Peripheral vascular disease (claudication or ischemic gangrene) or cerebral vascular disease (i.e., transient ischemic attack or stroke) f Including pulmonary edema within 6 months of starting dialysis g Serious illness limiting life expectancy to less than 5 years b c
Table 2. Primary diagnosis of end-stage renal disease
Diabetes mellitus Glomerulonephritis Hypertension Polycystic kidney disease Miscellaneous
Caucasian (N ⫽ 3752)
Southeast Asian (N ⫽ 451)
South Asian (N ⫽ 322)
Blacks (N ⫽ 319)
27% 21% 11% 7% 34%
32% 32% 11% 1% 24%
38% 25% 11% 3% 23%
37% 23% 15% 5% 20%
there was a higher prevalence of cardiovascular disease, cigarette smokers (both ex-and current smokers), and malignancies in the white patients at start of dialysis compared with the other racial groups. Table 3 shows the major causes of deaths in the four racial groups, with and without censoring at renal transplantation. Overall, 2104 patients died during the study period, and 90.6% of these deaths occurred during dialysis. In general, cardiovascular disease complications account for about half of the deaths in all racial groups. With the exception of a lower percentage of death from withdrawal of treatment in the black patients, there was no significant difference in the percentages of death in each of the main causes of death between the racial groups. Figure 1 shows the Kaplan–Meier survival curves of the four racial groups, showing all deaths that occurred during dialysis. Caucasian patients appeared to have the lowest survival among the four racial groups. The overall crude mortality rate during dialysis in white patients compared with Southeast Asian, South Asian, and black patients was 211 versus 128, 120, and 118 deaths per 1000 patientyears on dialysis, respectively. Multivariate models predicting patient survival In the best fitting models, all main effects except for gender had substantial power to predict patient survival.
The models were stratified by cardiovascular disease, hypertension, and diabetes mellitus, which all had large effects on survival. For parameters estimated in the proportional hazards regression, age, dialysis start year, malignancy, and congestive heart failure had the largest effects (in models with or without interactions), all with P values much less than 0.0001. In addition, most of the predictor variables were involved in multiple two-way interactions, suggesting that these interactions identify a high-risk patient profile that no one variable can independently capture. Having chosen the best model predicting survival without using any race information, race indicators were added to the model, and interactions between race and the risk factors were identified by using stepwise selection. The final models included the main effects (including race), all of the two-way risk factor interactions identified in the forward selection, and all interactions between white race and the other risk factors. The main effects, such as age, gender, and previously identified interactions, were not allowed to leave the model even if they were no longer significant at P ⬍ 0.05. All twoway interactions that entered each reported model are listed in Tables 4 and 5. In a few cases, the P values for the two-way interactions are greater than 0.05. This is
1296
Pei et al: Racial differences in survival of patients on dialysis Table 3. Causes of death of study patients
Deaths during dialysisa Cardiovascularb Sepsis Withdrawal from treatment Malignancy Miscellaneous All deathsc Cardiovascularb Sepsis Withdrawal from treatment Malignancy Miscellaneous a b c
Caucasian (N ⫽ 1614)
Southeast Asian (N ⫽ 130)
South Asian (N ⫽ 74)
Blacks (N ⫽ 88)
49% 15% 13% 7% 16%
48% 15% 15% 6% 16%
53% 14% 10% 5% 18%
56% 15% 6% 7% 16%
Caucasian (N ⫽ 1787)
Southeast Asian (N ⫽ 142)
South Asian (N ⫽ 83)
Blacks (N ⫽ 92)
48% 15% 13% 7% 17%
47% 14% 13% 7% 19%
51% 15% 10% 5% 19%
57% 14% 5% 8% 16%
Excluding all deaths after renal transplantation in patients who received a renal allograft Myocardial infarction, peripheral or cerebral vascular disease Including all deaths during dialysis and after renal transplantation in patients who received a renal allograft
Fig. 1. Kaplan–Meier survival curves of patients from the four racial groups, with patient follow-up censored at renal transplantation. Symbols are: (———) white; (··········) South Asian; (– – – –) South East Asian; (- - - - -) black. Caucasian patients had a lower overall survival rate compared with the other three racial groups (P ⬍ 0.0001).
because the best model, which did not include the race indicator, was founded before adding race and interactions with other variables. Of interest, a significant interaction between race and cigarette smoking was also detected (P ⬍ 0.004; see also the next section). Relative hazards of death between racial groups From the multivariate analyses, we derived pairwise relative hazards of death between the four racial groups. There was no evidence that the relative hazard of death associated with race changed with time. We found that white patients had an increased hazard of death during dialysis when compared with the three other racial groups (Table 4). Since a significant interaction between race and cigarette smoking had been detected, we also examined the relative hazards of death by smoking sta-
tus. We found that white smokers had a significantly increased hazard of death compared with smokers from the three other racial groups. For nonsmokers, white patients had a smaller but also increased hazard of death compared with patients from Southeast Asia and a trend for increased hazards of death compared with the South Asian and black patients (Table 4). Because the sample size in the non-white racial groups was relatively small, we were unable to determine whether there was any difference in the relative hazards of death between these racial groups (data not shown). We also addressed a potential concern that our findings might have been attributed to a bias in which a higher proportion of healthy Caucasians relative to the other racial groups was selected for renal transplantation. Such a bias could have led to a larger pool of
1297
Pei et al: Racial differences in survival of patients on dialysis Table 4. Relative hazards of death in study patients by racial groups Overall Death censored at renal transplantationa Caucasian vs. Southeast Asian Caucasian vs. South Asian Caucasian vs. black Caucasian vs. all Death post-transplantationa,b Caucasian vs. Southeast Asian Caucasian vs. South Asian Caucasian vs. black Caucasian vs. all
Nonsmoker
1.63 1.36 1.34 1.47
(1.36–1.97) (1.07–1.73) (1.07–1.67) (1.29–1.68)
1.12 1.04 2.44 1.30
(0.62–2.04) (0.53–2.04) (0.90–6.61) (0.85–1.99)
1.51 1.24 1.16 1.34
(1.24–1.83) (0.97–1.58) (0.92–1.47) (1.16–1.55)
Smoker 2.64 2.16 2.04 2.26
(1.75–3.98) (1.45–3.23) (1.39–2.97) (1.60–3.18)
a Relative hazards (95% CI) from Cox proportional hazards analysis stratified for cardiovascular disease, diabetes mellitus, and hypertension at start of dialysis and adjusted for all the main effects (age, gender, dialysis modality, dialysis start-year, serious illness at start of dialysis, cigarette smoking, malignancy, congestive heart failure, chronic obstructive lung disease, NRISK) and the following two-way interactions, all with P ⬍ 0.05 (hypertension ⫻ smoking; age ⫻ diabetes mellitus; diabetes mellitus ⫻ smoking; cardiovascular disease ⫻ malignancy; age ⫻ chronic obstructive lung disease; age ⫻ serious illness at start of dialysis; age ⫻ malignancy; age ⫻ NRISK; dialysis modality ⫻ serious illness at start of dialysis; dialysis start year ⫻ smoking; serious illness at start of dialysis ⫻ NRISK). b Deaths post-transplantation only in all patients who received a renal allograft. Analysis is stratified for cardiovascular disease, diabetes mellitus, and hypertension at start of dialysis, and adjusted for age, gender, transplant year, and malignancy.
Table 5. Relative hazardsa of death in study patients with or without clinical cardiovascular disease at initiation of dialysis Death censored at renal transplantation c Presented with cardiovascular disease
b
Presented without cardiovascular disease Caucasian Caucasian Caucasian Caucasian
vs. vs. vs. vs.
Southeast Asian South Asian black all
1.79 1.33 1.13 1.50
(0.43–2.26) (0.94–1.88) (0.85–1.51) (1.26–1.79)
Nonsmoker 1.20 1.27 1.33 1.26
(0.86–1.68) (0.90–1.80) (0.91–1.93) (1.00–1.59)
Smoker 2.39 2.53 2.64 2.53
(1.42–4.02) (1.49–4.28) (1.63–4.26) (1.62–3.92)
a Relative hazards (95% CI) from Cox proportional hazards analysis stratified for diabetes mellitus and hypertension at start of dialysis, and adjusted for all covariates that predict survival b For patients without cardiovascular disease at start of dialysis, the analyses were also adjusted for all the main effects as above, except NRISK and the following two-way interactions (age ⫻ diabetes mellitus, P ⫽ 0.00001; age ⫻ malignancy, P ⬍ 0.00001; hypertension ⫻ smoking, P ⬍ 0.005; age ⫻ serious illness at start of dialysis, P ⬍ 0.006; diabetes mellitus ⫻ malignancy, P ⬍ 0.01; dialysis modality ⫻ chronic obstructive lung disease, P ⬍ 0.05; hypertension ⫻ gender, P ⬍ 0.05; age ⫻ chronic obstructive lung disease, P ⬍ 0.05; diabetes mellitus ⫻ dialysis start-year, P ⬍ 0.01; dialysis start-year ⫻ smoking, P ⬍ 0.05; hypertension ⫻ chronic obstructive lung disease, P ⬍ 0.05) c For patients with cardiovascular disease at start of dialysis, the analyses were also adjusted for all of the main effects (age, gender, dialysis modality, dialysis start-year, serious illness at start of dialysis, cigarette smoking, malignancy, congestive heart failure, chronic obstructive lung disease, NRISK) and the following twoway interactions (dialysis modality ⫻ serious illness at start of dialysis, P ⬍ 0.00001; diabetes mellitus ⫻ chronic obstructive lung disease, P ⬍ 0.025; chronic obstructive lung disease ⫻ congestive heart failure, P ⬍ 0.02; hypertension ⫻ NRISK, P ⫽ 0.078).
sicker white patients on dialysis, thus accounting for the increased relative hazard of death among whites. A model that included a time-dependent covariate for transplant status did not detect a significant change in hazard with transplant for any racial group. Also, the relative hazards of death post-transplantation in our white patients were at or above unity compared with the other racial groups (Table 4). Taken together with the finding that only 9.4% of all deaths occurred after transplantation (Results section), these data provide strong evidence against a race-specific transplantation bias as an explanation for the increased relative hazards of death in our white patients. To examine the relationship between cigarette smoking and underlying cardiovascular disease to the overall risk of death further, a subgroup analysis was performed to compare the hazards of death between the four racial groups, with and without symptomatic cardiovascular
disease at the start of dialysis. Table 5 shows that in patients with symptomatic cardiovascular disease, white smokers had a significantly increased hazard of death compared with smokers from the other three racial groups. In contrast, for nonsmokers with or without symptomatic cardiovascular disease, there was a smaller and often nonsignificant trend for increased hazards of death in the white patients compared with the other racial groups. However, the overall relative hazard of death for Caucasians versus all other racial groups did not differ from the same racial comparison among patients who presented without cardiovascular disease (that is, 1.47 vs. 1.50, respectively; Tables 4 and 5). DISCUSSION Using a database that reflects the regional experience of a universal healthcare treatment system for ESRD
1298
Pei et al: Racial differences in survival of patients on dialysis
and a unique mixture of patients with different racial origins [5, 6], we have documented an increased mortality risk during dialysis in white patients compared with three other racial groups, after adjusting for differences in their comorbid risk factors. We have also excluded a potential concern that our findings might have been attributed to a bias in which a higher proportion of healthy whites were selected for transplantation, thus leaving a relatively larger pool of sicker white patients on dialysis. Atherosclerosis is a frequent and important cause of morbidity and mortality for patients with ESRD [10–16]. The notion that atherosclerosis is accelerated in the uremic milieu has been suggested [12–14] and is supported by the documented 15- to 20-fold increase in cardiovascular mortality in dialysis patients compared with ageand sex-matched populations without renal disease [10, 11, 15]. Thus, it is not surprising that cardiovascular disease accounts for 50% of all deaths in the patients with ESRD [11, 15, 16]. However, while a higher prevalence of cardiovascular disease at the start of dialysis was noted in whites compared with other racial groups (Table 1), this did not explain the observed differences in survival. One limitation of the current study was that only historic documentation of comorbid factors was used, and quantitative measures for the severity of these factors were not available. This may potentially limit our ability to detect whether the observed mortality differences could be related to different cardiovascular disease burden between the racial groups. Future studies that incorporate quantitative measures of cardiovascular disease severity are needed to clarify this issue further. Caucasian smokers with symptomatic cardiovascular disease at the start of dialysis constitute a very highrisk group for increased mortality. The basis for this observation is also presently unknown. Another limitation of the current study is that quantitative data on cigarette smoking in the study patients were not available. Thus, we cannot exclude the possibility that the mortality differences might have been related to racial differences in the cumulative dose of cigarette smoking, with the whites being the heaviest smokers. However, in a recent study of 144 healthy subjects without hypertension, hyperlipidemia, and diabetes mellitus, white smokers were found to have an increased impairment of arterial endothelial function compared with their Chinese counterparts when the dose of cigarette smoked was controlled [17]. These data suggest the possibility for a race-specific differential susceptibility to cardiovascular disease complications that may be related to environmental and/or genetic factors [17–21]. For example, racial differences in environmental factors such as dietary fat and antioxidant intake could have an important impact on cardiovascular disease mortality between the racial groups. However, comprehensive studies addressing the role of environmental factors on multiracial
cardiovascular disease burden are not presently available. Similarly, racial differences in genetic susceptibility factors may underlie the biologic variability of human disease traits such as cardiovascular disease complications [18]. In this context, a recent study has documented pharmacogenetic differences in nicotine metabolism between black and white smokers and suggests a differential risk for addiction and other medical complications of cigarette smoking between these racial groups [20, 21]. In the current study, nonsmoking white patients also had an increased hazard of death compared with Southeast Asian patients and a trend for increased hazards of death compared with South Asian and black patients (Table 4). Moreover, the overall hazard of death for whites versus all other racial groups did not differ from the same comparison among patients without symptomatic cardiovascular disease at the start of dialysis (that is, 1.47 vs. 1.5, respectively). These findings suggest that additional risk factors other than cigarette smoking and cardiovascular disease may influence survival. Of interest, dialysis dose has been suggested to be an important determinant for both morbidity and mortality in uremic patients [22, 23]. Furthermore, a reduced dialysis dose has been shown to be associated with increased mortality across all causes of death [24]. Since Southeast Asian and South Asian patients have a smaller body size and receive a higher average dialysis dose compared with Caucasian and black patients (abstract; Pei et al, J Am Soc Nephrol 8:208, 1997), racial differences in the dialysis dose could be an important determinant for the observed findings. On the other hand, a large registry study has failed to document that dialysis dose influences patient survival [25]. In addition to dialysis dose, the influence of race-specific differences in diet and lifestyle may be also important. All of these factors remain to be explored in future studies. In conclusion, we have documented an increased mortality risk in Caucasian patients with ESRD compared with three other racial groups, after adjusting for differences in their comorbid risk factors. At this time, we are unable to determine whether these mortality differences may be in part related to differences in the cardiovascular disease burden between these racial groups. However, white smokers with symptomatic cardiovascular disease at the start of dialysis constitute a very high-risk group for increased mortality compared with smokers from other racial groups. While our study has focused on uremic patients, racial differences in survival have been documented in the general population as well [26]. It is likely that both genetic and environmental factors may underlie these race-specific differences, but very little is currently known about these factors. Future studies to elucidate these race-specific, gene-environmental determinants of health and disease are greatly needed.
Pei et al: Racial differences in survival of patients on dialysis
ACKNOWLEDGMENTS We thank Dr. Stanley Fenton for his support of the Toronto Regional Dialysis Registry.
13.
Reprint requests to York Pei, M.D., Division of Nephrology, University Health Network, 13 EN-228, 200 Elizabeth Street, Toronto, Ontario, Canada, M5G 2C4. E-mail: [email protected]
14.
15.
REFERENCES 1. Agodoa L, Eggers P: Renal replacement therapy in the United States: Data from the United States Renal Data System. Am J Kidney Dis 25:119–133, 1995 2. USRDS 1997 Annual Data Report: Patient mortality and survival. Am J Kidney Dis 30:S91–S96, 1997 3. Mallick N, Jones E, Selwood N: The European (European Dialysis and Transplantation Association-European Renal Association) Registry. Am J Kidney Dis 25:176–187, 1995 4. Teraoka S, Toma H, Nihei H, Ota K, Babazono T, Ishikawa I, Shinoda A, Maeda K, Koshikawa S, Takahashi T, Sonoda T: Current status of renal replacement therapy in Japan. Am J Kidney Dis 25:151–164, 1995 5. Fenton S, Desmeules M, Copleston P, Arbus G, Froment D, Jeffrey J, Kjellstrand C: Renal replacement therapy in Canada: A report from the Canadian Organ Replacement Registry. Am J Kidney Dis 25:134–150, 1995 6. Statistics Canada: Ethnic origin, in 1991 Census Canada Ottawa: Industry, Science and Technology Canada (Catalogue Number 93– 315), 1993, pp 44–59 7. Kaplan E, Meier P: Nonparametric estimation from incomplete observations. J Am Stat Assoc 53:457–481, 1958 8. Armitage P, Berry G: Survival analysis, in Statistical Methods in Medical Research (2nd ed), Oxford, Blackwell Scientific Publications, 1987, pp 421–439 9. Therneau TM: Extending the Cox model, in Technical Report Number 58, pp 1–64, 1996, Rochester, Section of Biostatistics, Mayo Clinic 10. Wheeler D: Cardiovascular disease in patients with chronic renal failure. Lancet 348:1673–1674, 1996 11. London G, Drueke T: Atherosclerosis and arteriosclerosis in chronic renal failure. Kidney Int 51:1678–1695, 1997 12. Lindner A, Charra B, Sherrard D, Scribner B: Accelerated
16. 17.
18. 19.
20. 21. 22. 23.
24.
25. 26.
1299
atherosclerosis in prolonged maintenance hemodialysis. N Engl J Med 290:697–701, 1974 Ibels L, Alfrey A, Hufer W, Craswell P, Anderson J, Weil R: Arterial calcification and pathology in uremic patients undergoing dialysis. Am J Med 66:790–796, 1979 Kawagishi T, Nishizawa Y, Konishi T, Kawasaki K, Emoto M, Shoji T, Tabata T, Inoue T, Morii H: High-resolution ultrasonography in evaluation of atherosclerosis in uremia. Kidney Int 48:820– 826, 1995 Raine A, Margreiter R, Brunner F, Ehrich J, Geerlings W, Landais P, Mallick N, Selwood N, Tufveson G, Valderrabano F. Report on management of renal failure in Europe, XXII, 1991. Nephrol Dial Transplant 2:15–23, 1992 Parfrey P, Foley R, Harnett J, Kent G, Murray D, Barre P: Outcome and risk factors of ischemic heart disease in chronic uremia. Kidney Int 49:1428–1434, 1996 Woo K, Robinson J, Chook P, Adams M, Yip G, Mai Z, Lam C, Sorensen K, Deanfield J, Celermajer D: Differences in the effect of cigarette smoking on endothelial function in Chinese and white adults. Ann Intern Med 127:372–375, 1997 Schriver M: Ethnic variation as a key to the biology of human disease. Ann Intern Med 127:401–403, 1997 Wang X, Sim A, Badenhop, McCredie M, Wilken D: A smokingdependent risk of coronary artery disease associated with a polymorphism of the endothelial nitric oxide synthase gene. Nat Med 2:41–45, 1996 Sellers E: Pharmacogenetics and ethnoracial differences in smoking. JAMA 280:179–180, 1998 Perez-Stable E, Herrera B, Jacob P III, Benowitz N: Nicotine metabolism and intake in black and white smokers. JAMA 280: 152–156, 1988 Lowrie E, Laird N, Parker T, Sargent J: Effect of the hemodialysis prescription on patient morbidity. N Engl J Med 305:1176–1181, 1981 Owen W, Lew N, Liu Y, Lowrie E, Lazarus M: The urea reduction ratio and serum albumin concentration as predictors of mortality in patients undergoing hemodialysis. N Engl J Med 329:1001– 1006, 1993 Bloembergen W, Stannard D, Port F, Wolfe R, Pugh J, Jones C, Greer JW, Golper TA, Held PJ: Relationship of dose of hemodialysis and cause-specific mortality. Kidney Int 50:557–565, 1996 Owen W, Chertow G, Lazarus J, Lowrie E: Dose of hemodialysis and survival: Differences by race and sex. JAMA 280:1764–1768, 1998 Hahn R, Eberhardt S: Life expectancy in four U.S. racial/ethnic population. Epidemiology 6:350–355, 1995