Association of Physiological Reserve Obtained from Cardiopulmonary Exercise Testing and Frailty with All-Cause Mortality in Patients on Hemodialysis

Visual Abstract

Abstract

Key Points

• This study analyzed the prognostic value of reserves obtained from cardiopulmonary exercise testing in patients on hemodialysis.

• Each reserve (cardiac, autonomic, and muscular) was associated with mortality, with muscle reserve having the highest prognostic accuracy.

• Patients with physical frailty also had a fairly good prognosis if their physiological reserve is preserved by exercise testing.

Background

Potential impairment of exercise capacity is prevalent even in patients undergoing hemodialysis without frailty. Cardiopulmonary exercise testing (CPET) can detect physiological reserves, such as cardiopulmonary, muscle, and autonomic function. We hypothesized that these indices could accurately determine the prognosis of patients on hemodialysis and analyzed them on the basis of their relationship to frailty.

Methods

In this two-center prospective cohort study of patients on hemodialysis from Japan, patients underwent CPET and physical assessment to evaluate peak oxygen uptake (peak VO₂, indicator of exercise capacity), peak work rate (WR, indicator of muscle function), ventilatory equivalent for carbon dioxide (VE/VCO₂) slope (indicator of cardiac reserve), heart rate reserve (indicator of chronotropic incompetence), and frailty phenotype. Survival was followed up for up to 5 years.

Results

Data from 189 patients (median [interquartile range] age: 71 [62–77] years) were analyzed. All CPET indicators showed a consistent nonlinear relationship with all-cause mortality after adjustment: for peak VO₂, hazard ratio (HR), 0.79 (95% confidence interval [CI], 0.71 to 0.88), P < 0.001; for peak WR, HR, 0.95 (95% CI, 0.93 to 0.97), P < 0.001; for VE/VCO₂ slope, HR, 1.09 (95% CI, 1.05 to 1.13), P < 0.001; and for heart rate reserve, HR, 0.96 (95% CI, 0.93 to 0.99), P = 0.02. Frailty phenotype was associated with mortality after adjustment (HR, 1.73 [95% CI, 1.06 to 2.81], P = 0.03); however, this association was not statistically significant in the model after adding peak VO₂ (P = 0.41). Furthermore, in both subgroups with and without frailty, CPET measures were significantly associated with mortality risk (peak VO₂, peak WR, and VE/VCO₂ slope: P < 0.05). The peak VO₂ (Δ area under the curve, 0.09; 95% CI, 0.02 to 0.16) or the peak WR (Δ area under the curve, 0.09; 95% CI, 0.02 to 0.15) most significantly improved the prognostic accuracy.

Conclusions

Results showed the fragile aspect of the frailty phenotype in the hemodialysis population and the superior ability of CPET to indicate death risk complementing that aspect.

Introduction

Exercise intolerance and physical frailty are prevalent in patients with ESKD^1–3 and are associated with poor prognosis.^4,5 Peak oxygen uptake (peak VO₂) from cardiopulmonary exercise testing (CPET), the most reliable indicator of a person's exercise capacity, is the greatest prognostic indicator in primary prevention populations and patients with heart disease.⁶ Moreover, CPET can acquire a variety of parameters indicative of the reserve capacity of each organ (such as cardiac, pulmonary, skeletal muscle, and autonomic function parameters), and the combination of these parameters can accurately discriminate the risk of death in patients with heart failure.⁷ However, to our knowledge, despite the high morbidity and mortality associated with cardiovascular disease in patients on hemodialysis, no studies have evaluated the prognostic value of each reserve capacity obtained from CPET.

By contrast, the prevalence of frailty, which is defined as a biological syndrome of decreased reserve and resistance to stressors,⁸ is much higher in the population with hemodialysis.⁵ Although exercise capacity and physical frailty are indicators of a person's physiological reserve, exercise intolerance (which is defined by a decrease in peak VO₂) has been reported to be frequently observed even in patients on hemodialysis without physical frailty.^1,9 Thus, regardless of the presence or absence of physical frailty, exercise capacity (as assessed by CPET) is likely to be an excellent indicator of the physiological reserve across the full range of disease stages in this population. Nevertheless, to the best of our knowledge, no study has assessed the prognostic value of physical frailty, considering the confounding effect of exercise capacity. This could result in an overestimation of the frailty phenotype and an underestimation of the CPET indices as a clinical indicator of mortality risk. We performed an analysis using prospectively enrolled, larger sample size data to determine the prognostic value of each physiological reserve obtained from the CPET and the revised prognostic value of physical frailty in people on hemodialysis.

Methods

Design and Population

In this two-center prospective cohort study, adult outpatients on hemodialysis were enrolled for CPET and physical function assessment at Kisen Hospital and Kisen Clinic, located in different parts of the same city in Japan, from August 2017 to March 2023. Patients who were ineligible for the exercise test were as follows: patients with physical dysfunction (patients who cannot walk alone or those who cannot perform ergometer exercise in the upright sitting position), those with contraindications to the exercise test,¹⁰ and those who disagree with the study participation. On the basis of the contraindications to the exercise test, patients with unstable health status immediately after the initiation of hemodialysis were excluded. In addition, patients who underwent CPET with nonachievement of the test protocol (criteria are explained in the next section) and the presence of cardiac pacemakers were excluded from the statistical analysis. Patients were assessed on the same day as the CPET for the five components of physical frailty to diagnose their frailty phenotype. Each patient's survival status was followed annually for up to 5 years. The survival status and cause of death were investigated from electronic medical records. This study was approved by the Ethics Review Board of Kisen Hospital (approval No. 202003-001) and was conducted in compliance with the Declaration of Helsinki. Written informed consent was obtained from all patients.

Assessment of the CPET Status

Patients underwent CPET using a bicycle ergometer (STB-3300, OG Giken Co., Ltd., Japan) on a day after hemodialysis treatment (which is the day without hemodialysis treatment). The gradual load protocol of the CPET was performed using a short-duration ramp test to the limit of tolerance.¹¹ The exercise tests of all participating patients were examined by two experienced cardiologists. The test was terminated when 85% of the age-predicted maximal heart rate (HR) was attained or when exercise could not be continued because of physical exhaustion. CPET was discontinued if the patient requested for its discontinuation or if the physician discontinued the test for medical reasons.¹⁰ If patients failed to achieve any of the criteria (respiratory gas exchange ratio ≥1.1, Borg scale ≥17, and ≥85% of the age-predicted maximal HR), the protocol was considered incomplete, and they were excluded from the analysis. Peak VO₂ was calculated as the average oxygen uptake obtained during the last 30 seconds of CPET (exhaled gas analyzer; Minato Medical Science Co., Ltd., Japan). We also recorded the work rate (WR) at peak load during the exercise tests. The peak WR obtained from CPET is an indicator of peripheral function that reflects leg strength and muscle endurance.¹² Moreover, the ventilatory equivalent for carbon dioxide (VE/VCO₂) slope, which is an index of ventilatory efficiency and is inversely proportional to cardiac output during exercise, was calculated.¹³ VE/VCO₂ slopes were calculated by regression of the concurrent values of VCO₂ and VE before onset of acidotic drive to ventilation.¹⁴ The peak HR was recorded during CPET, and the HR reserve, which is an indicator of chronotropic incompetence,¹⁵ was evaluated using the following equation: HR reserve = [peak HR−rest HR]/[220−age−rest HR].¹⁶

Assessment of Physical Frailty Phenotype

The frailty phenotype was defined by frailty components (weakness, slowness, shrinking, exhaustion, and low activity) on the basis of the Cardiovascular Health Study criteria.^8,17 Patients were defined as frail if they had three or more components, prefrail if they had one or two components, and robust if they had zero components.⁸ Handgrip strength was measured using a digital grip dynamometer (Grip D dynamometer; Takei Scientific Instruments Co., Ltd., Niigata, Japan), and below sex-specific cutoff points (men: <28 kg, women: <18 kg) was determined to be weak. Gait speed was measured on a 4-m flat course, with a 1-m runway before the course. Patients were instructed to walk at their usual speed. A gait speed <1.0 m/s was defined as slowness. Shrinking was defined as unintentional weight loss of ≥2 kg or ≥5% over the previous 6 months.^17,18 Exhaustion was measured as a positive answer to the following self-reported questions: “In the last 2 weeks, have you felt tired without a reason?”^17,19 Low activity was defined as a “less than once a week” response to the questions “Do you engage in moderate levels of physical exercise or sports aimed at health?” and “Do you engage in low levels of physical exercise aimed at health?”¹⁷ Physical frailty as defined by this method is a standardized method in Japan and has been well validated in various populations, particularly the elderly and patients with cardiac disease and CKD.^1,17,20–25

Other Characteristics

The patients' baseline characteristics were investigated from electronic medical records. The comorbidity index for patients with hemodialysis²⁶ and Geriatric Nutritional Risk Index (GNRI) were calculated (GNRI=1.489×Alb [g/L]+41.7×body wt [kg]/ideal body wt [kg]).²⁷ In addition, predialysis serum hemoglobin levels were recorded.

Statistical Analysis

Data analyses were performed using Statistical Package for the Social Sciences software (ver. 26; IBM, NY) and R Commander Version 2.7-1 (The R Foundation for Statistical Computing, Vienna, Austria). Quantitative data are expressed as median (interquartile range [IQR]: 25th and 75th percentiles). First, peak VO₂ values were categorized into tertiles to compare patient characteristics and analyzed using chi-squared test and Kruskal–Wallis test, as appropriate.

For the cumulative survival analysis, patients began accruing risk time when they completed the exercise test and were followed until death or censorship at the end of the available follow-up period. Cox proportional hazards regression models were used to calculate the adjusted hazard ratios (HRs) for all-cause mortality associated with CPET parameters, frailty phenotype, and the five components of frailty. In the survival analysis, all CPET parameters were treated as continuous variables. We fitted four multivariate models on the basis of previous studies associated with exercise capacity⁴ and physical frailty⁵ in this population: The basic model included demographic factors (age, sex, hemodialysis vintage); the traditional model included demographic factors, use of β-blockers, comorbidity indexes for patients on hemodialysis,²⁶ GNRI, C-reactive protein, and hemoglobin; the frailty model included the frailty phenotype in addition to the traditional model; and the peak VO₂ model included peak VO₂ in addition to the traditional model. We also calculated adjusted HRs for all-cause mortality associated with peak VO₂ in the same model using peak VO₂ not standardized by body weight as a sensitivity analysis. Multicollinearity was tested by investigating variance inflation factor, checking for variables above 10. Next, restricted cubic splines were used to detect the possible nonlinear dependency of the association between each CPET parameter as a continuous variable and all-cause mortality using five knots at prespecified locations according to the 5th, 27.5th, 50th, 72.5th, and 95th percentiles. Restricted cubic splines were constructed using R with the “rms” and “Hmisc” packages.

The cumulative incidence of all-cause mortality between the four groups were compared on the basis of the presence of peak VO₂ impairment (<14 ml/kg per minute^4,28) and physical frailty (nonfrailty: robust or prefrail; frailty: frail) using Kaplan–Meier analysis and log-rank test. A Cox proportional hazards regression model was used to calculate adjusted HRs for all-cause mortality between the four groups. The same multivariate model (traditional model) was used for this analysis. In addition, we performed a Cox proportional hazards regression model with the same multivariate model (traditional model) to calculate adjusted HRs for all-cause mortality related to the CPET status in each subgroup on the basis of the presence of physical frailty. The same multivariate model was used to create an interaction term between frailty and CPET indices, whose adjusted HRs were also calculated. For all survival analyses, log–log survival curves were examined to confirm that there was no deviation from the proportional hazard assumption. Finally, C-statistics adapted for censoring was used to compare the discriminatory ability of survival models.²⁹ In addition, cross-validation was also performed to assess the accuracy of the prediction model. Missing values for key variables were imputed with the mean value.

Results

Baseline Characteristics

Figure 1 shows the patient flow diagram. Of the 277 patients with hemodialysis who were prospectively enrolled, 76 were excluded (physical dysfunction n=23, unstable health n=5, dementia n=6, and refusal n=42) and 201 underwent CPET. Of these, 12 were excluded (failure to achieve testing protocols n=9, cardiac pacemaker n=3) and data from 189 eligible patients were finally analyzed. Two patients had missing values for physical function data (handgrip strength n=2, and gait speed n=1), although the other data were fully available. Table 1 summarizes the patients' characteristics and present comparisons of characteristics between peak VO₂ tertiles and frailty phenotypes. Approximately 99% of all patients were Japanese. Patients had a median (IQR) age and hemodialysis vintage of 71 (63–77) and 4.2 (1.3–7.7) years, respectively, and included 23.8% women; median peak VO₂ was 12.7 (10.1–16.1) ml/kg per minute; the prevalence of frailty was 48% in the prefrail and 43% in the frail. The category with lower peak VO₂ had older age, higher comorbidity index, higher prevalence of diabetes, ischemic heart disease, and physical frailty, and lower GNRI and albumin (P < 0.05). However, as the severity of the frailty phenotype was higher, the CPET index was worse (P < 0.05).

Figure 1.

Flow diagram of patients. CPET, cardiopulmonary exercise test.

Table 1.

Patients characteristics

Variables	Overall (n=189)	Peak VO2 Category (ml/kg per minute)			P Value	Frailty Phenotypes			P Value
		High n=61 (>14.6)	Middle n=64 (10.7–14.6)	Low n=64 (<10.7)		Robust (n=18)	Prefrail (n=90)	Frail (n=81)
Age, yr	71 (63–77)	65 (54–74)	70 (66–76)	75 (71–79)	<0.001	65 (58–72)	71 (60–77)	73 (65–78)	0.08
Female, No. (%)	45 (24)	16 (25)	13 (20)	16 (26)	0.71	7 (39)	20 (22)	18 (22)	0.29
Hemodialysis vintage, yr	4.2 (1.3–7.7)	3.5 (1.1–7.5)	4.4 (1.3–8.7)	4.3 (1.7–7.4)	0.67	2.6 (1.3–5.7)	3.8 (1.3–7.7)	4.6 (1.2–8.6)	0.32
Body mass index, kg/m2	22.3 (20.2–25.3)	22.2 (19.5–24.7)	22.6 (21.1–25.8)	22.0 (20.0–25.1)	0.29	22.6 (20.2–26.3)	22.5 (20.5–25.3)	22.0 (19.7–24.9)	0.44
Primary disease, No. (%)					0.001				0.12
Nephrosclerosis	30 (16)	8 (13)	11 (17)	11 (18)		0 (0)	13 (14)	17 (21)
CGN	44 (23)	23 (36)	17 (27)	4 (6)		9 (50)	18 (20)	17 (21)
Diabetic kidney disease	85 (45)	22 (34)	24 (38)	39 (64)		6 (33)	45 (50)	34 (42)
Other	18 (10)	10 (16)	6 (9)	2 (3)		2 (11)	10 (11)	6 (7)
Unknown	12 (6)	1 (2)	6 (9)	5 (3)		1 (6)	4 (4)	7 (9)
Diabetes mellitus, No. (%)	86 (46)	22 (34)	25 (39)	39 (64)	0.002	5 (28)	46 (51)	35 (43)	0.17
Ischemic heart disease, No. (%)	49 (26)	7 (11)	17 (27)	25 (41)	<0.001	3 (17)	23 (26)	23 (28)	0.59
Chronic heart failure, No. (%)	30 (16)	7(11)	8 (13)	15 (25)	0.08	1 (6)	22 (24)	7 (9)	0.008
β-blocker, No. (%)	52 (28)	16 (25)	18 (28)	18 (30)	0.85	4 (22)	26 (29)	22 (28)	0.84
Comorbidity index	2 (1–4)	2 (0–3)	2 (1–4)	4 (2–5)	<0.001	1 (0–3)	3 (1–4)	2 (1–4)	0.07
GNRI	92.3 (88.0–95.3)	92.3 (88.5–95.3)	92.4 (90.0–96.3)	90.7 (86.4–93.1)	0.008	93.8 (90.8–96.8)	92.1 (87.9–95.3)	92.3 (88.1–94.5)	0.24
Hemoglobin, g/dl	10.9 (10.2–11.8)	10.9 (10.2–12.2)	11.1 (10.4–11.8)	10.7 (10.0–11.3)	0.14	11.1 (10.1–11.6)	10.9 (10.1–11.7)	10.9 (10.4–11.9)	0.83
Albumin, g/dl	3.5 (3.3–3.7)	3.6 (3.4–3.7)	3.5 (3.3–3.7)	3.4 (3.2–3.6)	0.002	3.6 (3.4–3.9)	3.4 (3.3–3.7)	3.5 (3.3–3.7)	0.10
C-reactive protein, mg/dl	0.13 (0.05–0.32)	0.08 (0.04–0.24)	0.13 (0.06–0.32)	0.17 (0.06–0.45)	0.06	0.11 (0.02–0.29)	0.11 (0.05–0.26)	0.14 (0.06–0.41)	0.36
CPET indices
Exercise time, s	394 (334–484)	488 (421–563)	388 (360–431)	316 (276–389)	<0.001	461 (411–581)	406 (347–498)	364 (301–427)	<0.001
RER	1.12 (1.07–1.19)	1.16 (1.10–1.26)	1.12 (1.08–1.17)	1.11 (1.02–1.18)	0.001	1.16 (1.12–1.23)	1.13 (1.07–1.23)	1.11 (1.07–1.17)	0.03
Borg scale	17 (15–17)	17 (15–17)	17 (15–17)	17 (15–17)	0.49	17 (15–17)	17 (15–17)	17 (15–17)	0.61
VO2 at, ml/kg per minute	9.3 (7.8–10.7)	11.2 (10.2–12.5)	9.4 (8.6–10.0)	7.3 (6.5–7.9)	<0.001	10.9 (9.4–12.5)	9.4 (8.1–10.7)	8.8 (7.0–10.2)	<0.001
Peak VO2, ml/kg per minute	12.7 (10.1–16.1)	17.1 (16.1–18.4)	12.4 (11.7–13.5)	9.5 (8.1–10.1)	<0.001	17.2 (14.0–18.3)	13.2 (10.3–16.3)	11.2 (9.5–14.1)	<0.001
Peak WR, W	63 (51–80)	83 (70–99)	61 (57–70)	47 (39–56)	<0.001	84 (67–100)	64 (54–82)	59 (46–68)	<0.001
VE/VCO2 slope	33.5 (28.9–39.0)	29.8 (27.5–34.1)	33.3 (29.2–39.2)	38.0 (33.6–44.9)	<0.001	30.4 (26.5–33.3)	33.4 (28.1–38.5)	34.8 (29.7–40.4)	0.006
HR reserve, %	20 (15–27)	27 (22–34)	20 (15–25)	14 (11–19)	<0.001	30 (27–34)	21 (15–28)	18 (13–24)	<0.001
Frailty phenotype, No. (%)					<0.001				—
Robust	18 (10)	12 (19)	6 (10)	0 (0)		—	—	—
Prefrail	90 (48)	35 (55)	28 (44)	27 (44)		—	—	—
Frail	81 (43)	17 (27)	30 (47)	34 (56)		—	—	—
Frailty component, No. (%)
Exhaustion	72 (38)	19 (30)	26 (41)	27 (44)	0.22	0 (0)	13 (14)	59 (73)	<0.001
Shrinking	59 (31)	22 (34)	18 (29)	19 (29)	0.75	0 (0)	15 (17)	44 (54)	<0.001
Weakness	117 (62)	25 (39)	41 (64)	51 (84)	<0.001	0 (0)	46 (51)	71 (88)	<0.001
Slowness	112 (59)	22 (34)	42 (66)	48 (79)	<0.001	0 (0)	44 (49)	68 (84)	<0.001
Low activity	69 (37)	22 (34)	24 (38)	23 (38)	0.91	0 (0)	23 (26)	46 (57)	<0.001

Values are median (25th and 75th percentiles), unless otherwise indicated. AT, anaerobic threshold; CGN, chronic GN; CPET, cardiopulmonary exercise test; GNRI, Geriatric Nutritional Risk Index; HR, heart rate; peak VO₂, peak oxygen uptake; RER, respiratory exchange ratio; VE/VCO₂, ventilatory equivalent for carbon dioxide; WR, work rate.

Survival Time Analysis

The median (IQR) follow-up period was 3.4 (1.8–4.2) years, and four patients (2%) were lost to follow-up. During the follow-up period, 54 patients (29%) died of the following causes: cardiovascular conditions (n=20), cancer (n=1), cerebrovascular conditions (n=8), infection (n=12), unknown causes (n=4), and other causes (n=9). Figure 2 shows Kaplan–Meier analysis of all-cause mortality for CPET indices. For all CPET indices, the risk of death was higher in the worse measured category (P < 0.001). Table 2 presents the results of the Cox proportional hazards regression models showing the relationship between all-cause mortality and CPET indices, frailty phenotypes, and each frailty component. First, for CPET indices, peak VO₂, peak WR, VE/VCO₂ slope, and HR reserve were all consistently and significantly associated with a higher risk of all-cause mortality in the unadjusted (all CPET indices: P < 0.001), basic (all CPET indices: P < 0.001), and traditional (all CPET indices: P < 0.05) models. Furthermore, even in the frailty model in which the frailty phenotype was added as a covariate, CPET measures of peak VO₂ (P < 0.001), peak WR (P < 0.001), VE/VCO₂ slope (P < 0.001), and HR reserve (P = 0.009) consistently retained their relationships with all-cause mortality. Moreover, the significant association between all-cause mortality and peak VO₂ was consistent in sensitivity analyses using absolute values of peak VO₂ that were not standardized by body wt (Supplemental Table 1). Figure 3 shows the nonlinear dependency of the association between CPET indices and mortality adjusted for covariates. Nonlinear associations with all-cause mortality were identified for all CPET indices, including peak VO₂ (P < 0.001), peak WR (P < 0.001), VE/VCO₂ slope (P < 0.001), and HR reserve (P < 0.001).

Figure 2.

Kaplan–Meier analysis of all-cause mortality for CPET indices. (A) Peak VO₂. (B) VE/VCO₂ slope. (C) Peak WR. (D) HR reserve. Each CPET index was categorized into tertiles. HR, heart rate; peak VO₂, peak oxygen uptake; VE/VCO₂, ventilatory equivalent for carbon dioxide; WR, work rate.

Table 2.

Adjusted relationship of cardiopulmonary exercise testing indices and frailty with all-cause mortality

Explanatory Variables	Unadjusted Model		Basic Model		Traditional Model		Frailty or Peak VO2 Model
	P Value	HR (95% CI)	P Value	HR (95% CI)	P Value	HR (95% CI)	P Value	HR (95% CI)
CPET indices							Frailty model
Peak VO2	<0.001	0.75 (0.68 to 0.83)	<0.001	0.76 (0.67 to 0.84)	<0.001	0.79 (0.71 to 0.88)	<0.001	0.80 (0.71 to 0.89)
Peak WR	<0.001	0.95 (0.94 to 0.97)	<0.001	0.94 (0.92 to 0.96)	<0.001	0.94 (0.92 to 0.97)	<0.001	0.94 (0.92 to 0.97)
VE/VCO2 slope	<0.001	1.10 (1.06 to 1.14)	<0.001	1.10 (1.06 to 1.14)	<0.001	1.09 (1.05 to 1.14)	<0.001	1.09 (1.05 to 1.13)
HR reserve	<0.001	0.92 (0.89 to 0.96)	<0.001	0.93 (0.89 to 0.96)	0.02	0.94 (0.90 to 0.98)	0.009	0.95 (0.91 to 0.99)

Frailty Indices							Peak VO2 Model
Frailty phenotype	0.02	1.71 (1.08 to 2.71)	0.06	1.57 (0.98 to 2.50)	0.04	1.65 (1.01 to 2.69)	0.63	1.14 (0.67 to 1.93)
Frailty component
Exhaustion	0.65	1.14 (0.66 to 2.00)	0.74	1.10 (0.63 to 1.91)	0.93	0.98 (0.55 to 1.74)	0.24	0.70 (0.39 to 1.27)
Shrinking	0.50	0.81 (0.44 to 1.50)	0.46	0.79 (0.43 to 1.47)	0.41	0.77 (0.42 to 1.43)	0.45	0.79 (0.42 to 1.46)
Weakness	0.03	2.06 (1.06 to 4.01)	0.17	1.64 (0.81 to 3.34)	0.13	1.82 (0.85 to 3.90)	0.79	1.12 (0.50 to 2.51)
Slowness	0.02	2.13 (1.12 to 4.06)	0.10	1.77 (0.90 to 3.48)	0.23	1.53 (0.77 to 3.04)	0.99	1.00 (0.50 to 2.02)
Low activity	0.11	155 (0.90 to 2.68)	0.07	1.67 (0.96 to 2.90)	0.02	2.00 (1.14 to 3.53)	0.04	1.78 (1.01 to 3.12)

Basic model: adjusted for age, sex, hemodialysis vintage. Traditional model: adjusted for age, sex, hemodialysis vintage, β-blocker, comorbidity index, Geriatric Nutritional Risk Index, C-reactive protein, and hemoglobin. Frailty model: adjusted for frailty phenotype in addition to the traditional model. Peak oxygen uptake model: adjusted for peak oxygen uptake in addition to the traditional model; gait speed and handgrip strength were analyzed as continuous variables. CI, confidence interval; CPET, cardiopulmonary exercise test; HR, heart rate; HR, hazard ratio; peak VO₂, peak oxygen uptake; WR, work rate; VE/VCO₂, ventilatory equivalent for carbon dioxide.

Figure 3.

Restricted cubic spline curve for all-cause mortality and CPET indices. Figures show a nonlinear association between all-cause mortality and each CPET parameter. The solid lines display the adjusted log HR, whereas the gray bands show the 95% confidence levels. The dashed lines show the reference line of HR. The models were adjusted for age, sex, hemodialysis vintage, β-blocker, comorbidity index, GNRI, C-reactive protein, hemoglobin, and frailty phenotype. GNRI, Geriatric Nutritional Risk Index; HR, hazard ratio.

Next, the results for physical frailty are presented. The frailty phenotype was significantly associated with all-cause mortality in both the unadjusted (P = 0.02) and traditional (P = 0.04) models. However, the peak VO₂ model, in which the peak VO₂ was added to the traditional model, did not retain significant relationships (P = 0.65). For each frailty component, weakness (P = 0.03) and slowness (P = 0.02) were associated with all-cause mortality in the unadjusted model. Moreover, only low activity retained an association with all-cause mortality in the peak VO₂ model (P = 0.04). No association with all-cause mortality was found for the other frailty components (traditional model: exhaustion P = 0.93, shrinking P = 0.41, weakness P = 0.13, slowness P = 0.23).

Survival Time Analysis in Four Groups Based on Peak VO₂ Impairment and Physical Frailty

Figure 4 (Kaplan–Meier analysis) and Table 3 (Cox models) show the comparison of mortality risk between the four groups on the basis of peak VO₂ impairment (<14 ml/kg per minute) and presence of physical frailty. Compared with the group without frail and peak VO₂ impairment, the risk of death was significantly higher in the group with peak VO₂ impairment alone (P = 0.02) and in the group with coexisting frail and peak VO₂ impairment (P = 0.004) after adjustment for confounding factors. However, no significant relationship to risk of death was observed in the frail-only group, which included a patient with no comorbid exercise capacity impairments, compared with the group without frail and peak VO₂ impairment (P = 0.73).

Figure 4.

Kaplan–Meier analysis of all-cause mortality in the four groups on the basis of peak VO₂ impairment and presence of physical frailty. Nonfrail includes robust and prefrail. Peak VO₂ impairment was defined as <14 ml/kg per minute.

Table 3.

Comparison of risk of all-cause mortality between the four groups on the basis of peak oxygen uptake impairment and presence of physical frailty

Explanatory Variables	Unadjusted Model		Basic Model		Traditional Model
	P Value	HR (95% CI)	P Value	HR (95% CI)	P Value	HR (95% CI)
Nonfrail and ≥14 ml/kg per minute	—	Reference	—	Reference	—	Reference
Frail alone	0.74	0.68 (0.07 to 6.52)	0.68	0.62 (0.06 to 5.94)	0.73	0.67 (0.07 to 6.52)
<14 ml/kg per minute alone	0.001	7.50 (2.24 to 25.16)	0.003	6.30 (1.84 to 21.62)	0.02	4.69 (1.31 to 16.82)
Frail and <14 ml/kg per minute	<0.001	8.94 (2.72 to 29.44)	0.001	7.42 (2.21 to 24.91)	0.004	5.98 (1.75 to 20.41)

Basic model: adjusted for age, sex, hemodialysis vintage. Traditional model: adjusted for age, sex, hemodialysis vintage, β-blocker, comorbidity index, Geriatric Nutritional Risk Index, C-reactive protein, and hemoglobin. Nonfrail includes robust and prefrail. Peak oxygen uptake impairment was defined as <14 ml/kg per minute. CI, confidence interval; CPET, cardiopulmonary exercise test; HR, hazard ratio; peak VO₂, peak oxygen uptake.

Results of the Cox proportional hazards regression models showed that the peak VO₂ (nonfrail: P = 0.01, frail: P = 0.003), peak WR (nonfrail: P = 0.005, frail: P < 0.001), and VE/VCO₂ slope (nonfrail: P = 0.003, frail: P = 0.002) were associated with all-cause mortality in all subgroups. The traditional model was fitted to this analysis (Figure 5). Meanwhile, the HR reserve was not associated with all-cause mortality in all subgroups (nonfrail: P = 0.13, frail: P = 0.06).

Figure 5.

Association analysis of all-cause mortality in subgroups on the basis of the presence of frailty. The models were adjusted for age, sex, hemodialysis vintage, β-blocker, comorbidity index, GNRI, C-reactive protein, and hemoglobin. CI, confidence interval.

Analysis of the C-Statistic Adapted for Censoring

We compared the discrimination of the Cox models using C-statistics (Table 4). When compared with the traditional model, the addition of the peak VO₂ (Δ area under the curve, 0.09; 95% confidence interval, 0.02 to 0.16) or peak WR (Δ area under the curve, 0.08; 95% confidence interval, 0.01 to 0.16) significantly improved the predictive accuracy for death. The addition of other indices associated with CPET and physical frailty did not improve the predictive accuracy for death. The cross-validation results confirmed the internal and external validity of all prediction models, indicating that the models have excellent prognostic accuracy (Supplemental Table 2).

Table 4.

Comparison of C-statistics of the adjusted Cox proportional hazard model for all-cause mortality

Models	AUC (95% CI)	P Value	Δ AUC (95% CI)	P Value
Traditional model	0.75 (0.67 to 0.82)	<0.001	Reference	—
+Peak VO2	0.84 (0.78 to 0.90)	<0.001	0.09 (0.02 to 0.16)	0.009
+Peak WR	0.83 (0.77 to 0.89)	<0.001	0.08 (0.01 to 0.16)	0.02
+VE/VCO2 slope	0.78 (0.71 to 0.86)	<0.001	0.03 (−0.02 to 0.09)	0.21
+HR reserve	0.78 (0.71 to 0.86)	<0.001	0.04 (−0.02 to 0.09)	0.16
+Frailty phenotype	0.77 (0.70 to 0.84)	<0.001	0.02 (−0.01 to 0.05)	0.22
Frailty component
+Exhaustion	0.75 (0.67 to 0.82)	<0.001	0.00 (−0.01 to 0.01)	0.89
+Shrinking	0.75 (0.68 to 0.83)	<0.001	0.01 (−0.01 to 0.02)	0.47
+Weakness	0.77 (0.70 to 0.84)	<0.001	0.02 (−0.01 to 0.06)	0.18
+Slowness	0.76 (0.68 to 0.83)	<0.001	0.01 (−0.01 to 0.03)	0.26
+Low activity	0.76 (0.69 to 0.83)	<0.001	0.02 (−0.02 to 0.05)	0.44

The model includes age, sex, hemodialysis vintage, β-blocker, comorbidity index, Geriatric Nutritional Risk Index, C-reactive protein, and hemoglobin; AUC, area under the curve; CI, confidence interval; HR, heart rate; peak VO₂, peak oxygen uptake; WR, work rate; VE/VCO₂, ventilatory equivalent for carbon dioxide.

Discussion

Our study conducted on patients on hemodialysis revealed some important novel findings. First, the peak VO₂ (a measure of a person's exercise capacity) and all physiological reserves obtained from the CPET, including the peak WR (an indicator of muscle strength and endurance¹²), VE/VCO₂ slope (an indicator of cardiac output during exercise¹³), and HR reserve (an indicator of chronotropic incompetence¹⁵), which are determinants of a person's exercise capacity, remained associated with a higher risk of all-cause mortality after adjustment for confounders in a multivariate model. In addition, the groups with peak VO₂ impairment alone and with coexisting frail and peak VO₂ impairment had a higher risk of death; by contrast, the group without peak VO₂ impairment had a good prognosis even in the presence of frailty. Furthermore, each CPET measure (except HR reserve) determined the risk of death distinctly in populations with and without physical frailty. These results show that each physiological reserve obtained from CPET are useful indicators of the physiological reserve across the full range of disease stages in this population, regardless of physical frailty. In addition, the fact that all measures obtained from CPET were consistently associated with a higher risk of death indicates that both peripheral (skeletal muscle) and cardiac factor reserves (which determine peak VO₂) are closely related to outcomes in this population. In particular, the peak VO₂ and peak WR (measures of muscle strength and endurance) were found to greatly improve the prognostic accuracy of traditional models by approximately 10%.

A previous study reported that the ability of the exercising skeletal muscles to extract oxygen is the predominant determinant of the exercise capacity of patients with CKD, followed by the heart's ability to generate the stroke volume and raise the HR.³⁰ Our results showed that after the peak VO₂, the peak WR (as determined by a person's muscle strength and endurance) had the most improved prognostic accuracy. This is consistent with previous findings that the decrease in peak VO₂ is most influenced by the exercise ability of skeletal muscles, indicating a substantial contribution to the peak VO₂ of skeletal muscle function, a peripheral factor in this population. Furthermore, our results showed that the VE/VCO₂ slope, which is influenced by higher cardiac output during exercise, was also closely associated with a higher risk of death in this population. A previous study reported that the cardiac functional reserve is latently impaired early, even in asymptomatic patients with CKD.³¹ The potentially impaired exercise tolerance associated with the impaired cardiac reserve in this stable patient with CKD is probably the result of uremic cardiomyopathy,³² which is frequent in this population. The VE/VCO₂ slope is probably an early diagnostic indicator of a potentially impaired cardiac reserve associated with uremic cardiomyopathy in this population.

A previous study demonstrated that the frailty phenotype of the hemodialysis population correlates well with the CPET indices.¹ Furthermore, the results of our study and those of a previous study⁵ showed a consistent relationship between the frailty phenotype and all-cause mortality in traditional prognostic models. Thus, the frailty phenotype has a certain level of utility as a risk stratification indicator that can be easily used in a clinical setting and can predict the peak VO₂. In patients with physical frailty on hemodialysis, a severe decline in exercise capacity is naturally observed.¹ In patients with manifestations of declining physical performance (such as physical frailty), a comprehensive lower extremity functional assessment, such as the Short Physical Performance Battery³³ (an index consisting of gait speed, lower extremity muscle strength, and balance function), accurately determines the risk of death.^34–36 This relationship between physical performance and risk of death is consistent even in CKD, including patients on hemodialysis.^37,38 Thus, even in the hemodialysis population, patients with physical frailty can likely be conveniently and accurately discriminated for the risk of death using a comprehensive physical performance battery. Conversely, exercise intolerance is frequently observed even in the active hemodialysis population without physical frailty.^1,9 Our results show that the peak VO₂ and other CPET indices accurately predicted the risk of death even in patients without physical frailty. Thus, in patients without physical frailty whose physical performance is preserved, CPET is the only assessment tool that can detect potential cardiopulmonary reserve impairment. Our findings reveal the partial limitations of the frailty phenotype in the hemodialysis population and the superior discrimination ability of the CPET indices for the risk of death to complement these limitations.

In our results, among the physical frailty components, only low activity was consistently associated with a higher risk of all-cause mortality. Patients on hemodialysis remain substantially less active than the general population with normal kidney function, and physical inactivity is strongly associated with higher risk of mortality.^39,40 Physical inactivity can induce a catabolic state, including reduced neuromuscular functioning, lower exercise tolerance, and lower cardiorespiratory fitness,⁴¹ which is the genesis of subsequent adverse health outcomes.^40–43 Physical inactivity in CKD lowers total energy expenditure and causes malnutrition owing to dietary deficiencies, leading to weight and skeletal muscle loss.⁴⁴ Physical inactivity is a critical factor that accelerates the vicious cycle of frailty and seems to reflect an important dimension distinct from exercise capacity that causes serious outcomes in this population.

On the basis of our findings, CPET indices recommended to be used as one of the most important intermediate outcomes for therapeutic interventions to improve serious outcomes in this population. However, CPET is not common in this field, and in many cases, performing CPET on all patients in a clinical setting is difficult. CPET is a limited tool that can detect potential impaired reserve related to cardiopulmonary or skeletal muscle function in patients who are asymptomatic with preserved physical function. Exercise interventions for patients with hemodialysis are well known to improve exercise capacity and physical function⁴⁵ and have recently been reported to lower the risk of hospitalization or death.⁴⁶ Moreover, frailty in patients on hemodialysis is associated with higher health care utilization costs,⁴⁷ meanwhile exercise intervention for this population has been shown to be a cost-effective treatment.⁴⁸ Thus, detecting impaired reserve in patients on hemodialysis who were asymptomatic, motivating them, and providing early preventive therapeutic intervention seems to be effective in improving patient outcomes and reducing health care expenditures, beyond the effort and expense of CPET.

Nevertheless, our study had several limitations. The small sample size of robust (nonfrail) frailty phenotypes may have led to uncertain estimates of frailty phenotypes for the risk of death and an underestimation of its association. Alternatively, the strength of the CPET indices is that they accurately determined the risk of death even in such a population and their estimates were stable. In addition, the patients included a large number of older people and a small number of obese people. Therefore, notably, generalization of our results may be limited to a relatively older population without obesity. Moreover, the effects of residual confounding can attenuate the magnitude of the true association.

In conclusion, our results reveal the fragile aspect of the frailty phenotype in the hemodialysis population and the superior ability of physiological reserve identified by CPET to determine the risk of death complementing that aspect.

Supplementary Material

cjasn-20-420-s001.pdf ^{(1.5MB, pdf)}

cjasn-20-420-s002.xlsx ^{(58.7KB, xlsx)}

cjasn-20-420-s003.pdf ^{(160.6KB, pdf)}

Disclosures

Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/CJN/C132.

Funding

J. Nakata: Japan Society for the Promotion of Science (21K11199).

Author Contributions

Conceptualization: Junichiro Nakata, Yusuke Suzuki, Akimi Uehata, Naoto Usui.

Data curation: Sho Kojima, Naoto Usui.

Formal analysis: Shuji Ando, Sho Kojima, Naoto Usui.

Funding acquisition: Junichiro Nakata.

Investigation: Hideki Hisadome, Akihito Inatsu, Sho Kojima, Yuki Nishiyama, Akimi Uehata, Naoto Usui.

Methodology: Junichiro Nakata, Masakazu Saitoh, Yusuke Suzuki, Naoto Usui.

Project administration: Yusuke Suzuki, Akimi Uehata.

Resources: Akimi Uehata.

Supervision: Junichiro Nakata.

Validation: Junichiro Nakata, Masakazu Saitoh.

Visualization: Naoto Usui.

Writing – original draft: Junichiro Nakata, Naoto Usui.

Writing – review & editing: Shuji Ando, Hideki Hisadome, Akihito Inatsu, Sho Kojima, Yuki Nishiyama, Masakazu Saitoh, Yusuke Suzuki, Akimi Uehata.

Data Sharing Statement

All data are included in the manuscript and/or supporting information.

Supplemental Material

This article contains the following supplemental material online at http://links.lww.com/CJN/C130, http://links.lww.com/CJN/C131.

Supplemental Table 1. Results of the Cox proportional hazards model for the association between absolute peak VO₂ (unstandardized by body weight) and all-cause mortality.

Supplemental Table 2. Cross-validation results for C-statistics of the adjusted Cox proportional hazard model for mortality.

Supplemental Spreadsheet: Underlying Dataset.