Background

Exercise capacity is reduced in end-stage renal disease (ESRD). Exercise requires the integrated function of multiple vital organs, and low exercise capacity is an independent predictor of mortality in a number of clinical populations. We analyzed the value of exercise capacity, characterized as peak oxygen uptake (VO₂), for predicting survival in a cohort of 175 hemodialysis patients over a median follow-up of 39 months.

Methods

Survival status was determined for 175 ESRD patients who had participated in previous studies for which peak VO₂ and other clinical data had been determined. Chi-square and Kaplan-Meier survival analyses were performed, and a minimal model of factors related to mortality was developed by Cox multiple regression.

Results

There were 23 deaths during the follow-up period. Peak VO₂ (>17.5 mL/min/kg) was a powerful predictor of survival (P = 0.009 by Kaplan-Meier). Age (<65 years), dialysis vintage (<39 months), pulse pressure (<54 mm Hg), and absence of diagnoses of diabetes or heart failure were also associated with better survival on univariate analyses. On multivariate analysis peak VO₂ contributed significantly to the minimal explanatory model relating clinical variables to mortality (overall ²= 25.5, P = 0.00001).

Conclusion

Among these ambulatory ESRD patients, peak VO₂ was a stronger predictor of survival than many traditional prognostic variables, some of which are subject to ceiling effects. Exercise capacity may thus provide incremental prognostic information concerning healthier ESRD patients. Because peak VO₂ may be modified by exercise training, the potential of exercise as an intervention to improve survival is suggested.

METHODS

Study subjects

The analysis made use of a database of clinical, demographic, and exercise variables for patients with ESRD who participated in two clinical trials of L-Carnitine (Carnitor®) supplementation in chronic hemodialysis (CHIEF studies)²⁰. Key inclusion criteria for those studies included ESRD on thrice weekly hemodialysis for at least 6 months with Kt/V value of 1.2, clinical stability, and ability to perform incremental exercise testing to symptomatic maximum. Study participation was six months' duration, during which time peak VO₂ was determined at three-month intervals using a standardized cycle ergometry exercise testing protocol. Patients were excluded from study if they could not cooperate with testing or if an initial screening exercise test had to be terminated because of adverse reactions such as ischemic changes or arrhythmia on electrocardiogram, or pathologic blood pressure response. The present analysis is based on peak VO₂ measures determined on a baseline test performed on a separate day from the screening test.

Approximately three years after completion of the CHIEF studies, a questionnaire was mailed to each participating site to ascertain survival status and dates and causes of deaths as of October 1, 2001 for patients who had been enrolled in the trials. Ten of the original 12 study sites participated in the follow-up. Collection and use of these data for the present analysis was reviewed and approved by the institutional review board of the Harbor-UCLA Research and Education Institute.

Analyses

Patient characteristics as related to outcomes were summarized using descriptive statistics. Group values are reported as mean and SD unless otherwise noted. The primary outcome measure was survival during the follow-up period. Simple descriptive statistics were used to characterize survivors and non-survivors, who were compared using either analysis of variance (ANOVA) or chi-square analysis with Yate's correction where applicable. Survival status was analyzed as a function of exercise capacity, expressed as peak VO₂ in mL/kg/min, and of other selected clinical variables. Chi-square analysis was used to compare crude death rates and univariate survival analyses were performed using the log-rank test on survival curves created with the method of Kaplan-Meier. For these purposes stratification of the population for continuous variables was done using values previously utilized in survival analyses for this patient population (for age and serum albumin), or using the observed group median value (for peak VO₂, dialysis vintage, serum creatinine, blood pressures, and hemoglobin). Sensitivity analyses were also conducted for some of these variables, using different cut-off points. Survival data were not censored at the time of kidney transplantation. To control for covariance of exercise capacity with other variables recognized to have prognostic significance in this population, Cox's proportional-hazards multivariate regression was performed. All potentially relevant variables identified from Kaplan-Meier survival analyses (P < 0.15) were entered into the Cox model; a minimal explanatory model predicting mortality was then derived using a backward selection approach. All analyses were performed using Statistica v. 5.1 software (StatSoft, Inc., Tulsa, OK, USA).

RESULTS

Data were available for all patients from the 10 reporting study sites, representing 175 of 193 patients enrolled in the CHIEF studies. Survival status and dates of deaths were determined approximately 3.5 years after the initial baseline assessments were performed (mean follow up: 1242 273 days). There were a total of 23 deaths (13%) during the follow-up period. Cardiovascular causes were reported for seven deaths (including cardiac arrest, cardiomyopathy, and cerebral ischemia), infectious causes for seven (including sepsis, abscess, and pneumonia), and malignancy for two. Causes of the remaining seven deaths were reported as unknown. Forty-four patients received kidney transplants during the follow-up period, and two of the deaths occurred among these patients.

Subject characteristics

Baseline characteristics for all patients and for the subgroups of surviving and non-surviving patients are shown in Table 1. Patients averaged 46 13 years of age and had been on dialysis an average of 41 15 months at the time of the baseline evaluations. Group mean values for clinical variables reflect, in part, the entry criteria for the original study which included requirements of clinical stability, adequate dialysis as reflected in a Kt/V of at least 1.2, and treatment with erythropoietin in accordance with prevailing standards of clinical practice.

Table 1 - Patient characteristics.

Full table

Factors influencing survival

Crude death rates for patients stratified by selected clinical variables are shown in Table 2. Patients with a peak VO₂ above the group median value of 17.5 mL/min/kg had a significantly better crude survival rate than those with lower peak VO₂ Table 2; this finding was confirmed by univariate Kaplan-Meier analysis (P = 0.009) Figure 1. Similar results were obtained if peak VO₂ was stratified at 14 or 15 mL/min/kg (P = 0.0016 and 0.00064, respectively; data not shown). Patients younger than age 65, those with resting pulse pressure of less than 54, and those without a diagnosis of diabetes, or without a diagnosis of chronic heart failure, had a better crude survival than those who were older, had a higher pulse pressure, were diabetic, or suffered from chronic heart failure. There was a trend (P = 0.15) toward better crude survival for those whose serum albumin was at or above the group median value of 4.0 g/dL Table 2, and this trend was confirmed on univariate Kaplan-Meier analysis (P = 0.08) Figure 2. In contrast, hemoglobin and serum creatinine were not significantly related to mortality in this analysis Table 2, nor were race or gender (data not shown).

Figure 1.

Survival as a function of baseline value of peak exercise VO₂ for 175 ambulatory patients with end-stage renal disease. Patients with peak VO₂ above the median value of 17.5 mL/min/kg (interrupted line) had significantly better survival than those with lower values (solid line) (P = 0.009 by log-rank test).

Full figure and legend (23K)

Figure 2.

Survival as a function of baseline serum albumin for 175 ambulatory patients with end-stage renal disease. Serum albumin concentration at or above the group median value of 4.0 gm/dL (interrupted line) showed a trend toward better survival than those with lower values (solid line) (P = 0.08 by log-rank test).

Full figure and legend (28K)

Table 2 - Crude mortality rates for 175 patients with ESRD stratified according to selected clinical variables.

Full table

Multivariate analysis was somewhat limited by the relatively low number of deaths and the large number of variables available. Nonetheless, a highly significant minimal model (overall ²= 25.5, P = 0.00001) was developed that included age, peak VO₂, and the diagnosis of chronic heart failure as predictors of mortality Table 3.

Table 3 - Independent variables and minimal model of factors related to mortality in 175 patients with ESRD.

Full table

DISCUSSION

In this analysis of data from 175 ambulatory hemodialysis patients, peak VO₂ was a powerful predictor of survival status over an average follow-up of approximately 3.5 years. In a multivariate minimal explanatory model, the only additional factors significantly enhancing this predictive value were age and the presence or absence of chronic heart failure.

Individual variables related to prognosis in hemodialysis populations include functions of nutrition, inflammation, cardiovascular status, and comorbid medical conditions. Serum concentrations of albumin¹, prealbumin², and creatinine¹ all have positive relationships to survival, consistent with their relationship to protein balance and nutritional state. There are negative associations between inflammatory markers and survival^8,21, which may relate in part to the negative effect of inflammation on protein balance²² as well as pathogenic processes underlying vascular disease⁷. More distal measures of cardiovascular dysfunction, including left ventricular hypertrophy¹⁴ and pulse pressure¹², or chronic heart failure, are associated with mortality, as is the diagnoses of diabetes^10,23. Not surprisingly, many of these factors covary, reflecting both commonality of cause and physiologic interactions among the processes. Consistent with this, in the present analysis a number of variables including serum albumin, serum creatinine, and dialysis vintage, which were each significantly related to outcome on univariate analysis, were effectively incorporated into a smaller number of variables on mulitvariate regression.

Exercise capacity of healthy persons is intrinsically dependent on many of the same variables that predict survival in hemodialysis patients^24,25. Specifically, the capacity to transport oxygen is determined by cardiovascular function and hemoglobin concentration, and the capacity for utilization of oxygen is related to skeletal muscle mass and function, which are dependent on nutritional status. Both diabetes and hypertension, which may predict poor outcome in ESRD, are associated with reduced exercise capacity even in persons without renal failure^26,27. We have previously reported correlations between exercise capacity of hemodialysis patients and a number of these variables, including hemoglobin concentration, serum albumin, and the diagnosis of diabetes²⁸. It is thus reasonable to expect that peak exercise VO₂ would represent an integrated measure of multiple important prognostic variables in ESRD, and therefore would itself be related to outcome. Apart from serving indirectly as a marker of other prognostic factors, however, peak VO₂ directly reflects the capacity to increase cardiac output in response to the physiologic stress of exercise. As such, it is likely to correspond to the capacity to survive pathologic stresses related to acute illness or injury, as well.

The patients included in this analysis represent a healthy subset of the hemodialysis population, reflected in an overall mortality rate that was far lower than that reported for United States hemodialysis patients in general²⁹. By initial study inclusion criteria, the patients were judged to be clinically stable and able to perform maximal exercise testing. Non-ambulatory, frail, or unstable patients, who would be predicted to be at highest risk for mortality based on commonly used prognostic variables, are, thus, not represented. This may explain why serum albumin was a relatively weak predictor of survival in this study, despite being among the strongest reported prognostic variables in large ESRD cohorts¹. Patients who had findings of significant exercise-induced EKG abnormalities on screening exercise tests were also excluded from the study, although a substantial proportion of the observed mortality was nevertheless attributed to cardiovascular disease. The prognostic power of peak VO₂ in this relatively healthy subset of the dialysis population is thus of particular interest, as the mortality risk of these patients may be less readily predicted by other assessments due to threshold effects³⁰.

Exercise capacity has been recognized as an independent predictor of survival across a broad range of values within healthy populations^31,32,33,34. A similar relationship between exercise capacity and prognosis has been identified in certain clinical conditions, particularly cardiovascular diseases^19,35. This is particularly well-documented in chronic heart failure¹⁷, and has practical application in prioritization and timing of heart transplantation. In that setting, peak VO₂ values less than 14 mL/min/kg have been used as a criteria for heart transplant eligibility³⁶, based on observations that higher values predicted a two-year survival as good as that of coincident new transplant recipients. In neither the heart failure population¹⁷ nor among apparently healthy men³⁴ does there appear to be a ceiling above which the relationship between peak VO₂ and prognosis is lost, even after control for other commonly used clinical and demographic risk factors.

The number of deaths in the present study was low, limiting the statistical power of multivariate analyses to fully explore the relationships between peak VO₂, other clinical factors, and mortality. The highly significant relationship between exercise capacity and survival on univariate analysis, and its contribution to the multivariate model supports further investigation of this, however, particularly with respect to risk stratification of the healthiest and most functional ESRD patients.

Ultimately, the value of identifying predictors of survival or mortality lies in the identification of modifiable factors to target for intervention. It is noteworthy therefore that peak VO₂ may be increased by physical exercise training. Furthermore, it appears that training reduces mortality in the general population^37,38 and perhaps among patients with heart failure³⁹. Among patients with ESRD exercise training has been demonstrated to increase peak VO₂⁴⁰ and functional capacity⁴¹. It has also recently been reported⁴² that dialysis patients who identify themselves as sedentary have higher mortality than those who report being ambulatory. Whether sedentary behavior contributes causally to increased mortality in this setting, or is simply a marker of more extensive disease and poorer prognosis is not clear. Thus, it remains to be seen whether increasing peak VO₂ by exercise training might modify mortality rate in this population.

CONCLUSION

In this study peak VO₂ was found to be a highly significant predictor of survival among a stable ambulatory cohort of hemodialysis patients. This finding is consistent with other data demonstrating that nutritional status, cardiovascular factors, and exercise behavior are significant factors related to outcome in ESRD, and suggests that exercise testing may represent a useful approach to risk stratification among the most functional patients with ESRD. The potential for interventions targeting exercise function to further enhance survival in this population is deserving of further study.

References

References

Acknowledgments

The authors wish to thank the investigators of the CHIEF trials for providing data for this analysis. This project was supported in part by Sigma Tau Research, Inc.