Physical activity and risk of cardiovascular events and all-cause mortality among kidney transplant recipients

Abstract

Background

Insufficient physical activity (PA) may increase the risk of all-cause mortality and cardiovascular disease (CVD) morbidity and mortality among kidney transplant recipients (KTRs), but limited research is available. We examine the relationship between PA and the development of CVD events, CVD death and all-cause mortality among KTRs.

Methods

A total of 3050 KTRs enrolled in an international homocysteine-lowering randomized controlled trial were examined (38% female; mean age 51.8 ± 9.4 years; 75% white; 20% with prevalent CVD). PA was measured at baseline using a modified Yale Physical Activity Survey, divided into tertiles (T1, T2 and T3) from lowest to highest PA. Kaplan–Meier survival curves were used to graph the risk of events; Cox proportional hazards regression models examined the association of baseline PA levels with CVD events (e.g. stroke, myocardial infarction), CVD mortality and all-cause mortality over time.

Results

Participants were followed up to 2500 days (mean 3.7 ± 1.6 years). The cohort experienced 426 CVD events and 357 deaths. Fully adjusted models revealed that, compared to the lowest tertile of PA, the highest tertile experienced a significantly lower risk of CVD events {hazard ratio [HR] 0.76 [95% confidence interval (CI) 0.59–0.98]}, CVD mortality [HR 0.58 (95% CI 0.35–0.96)] and all-cause mortality [HR 0.76 (95% CI 0.59–0.98)]. Results were similar in unadjusted models.

Conclusions

PA was associated with a reduced risk of CVD events and all-cause mortality among KTRs. These observed associations in a large, international sample, even when controlling for traditional CVD risk factors, indicate the potential importance of PA in reducing CVD and death among KTRs.

Graphical Abstract

Graphical Abstract.

INTRODUCTION

Long-term outcomes among kidney transplant recipients (KTRs) have not improved over the last decades [1]. Cardiovascular disease (CVD) is the leading cause of death among KTRs [2] and KTRs have a higher prevalence of CVD and CVD-related mortality compared with the general population (e.g. the risk of CVD mortality is 10 times higher than the general population [3] and the risk of CVD events is 50 times higher than in the general population [4]). In addition, CVD events are more likely to be fatal among KTRs than in the general population. Both traditional and nontraditional risk factors are likely explanations for the elevated CVD morbidity and mortality [5, 6], although few research studies to date have examined this topic. Renal transplantation procedures are increasingly performed every year (with almost 20 000 renal transplantations performed in the USA alone in 2017) [7]; hence, examining the factors underlying CVD risk among KTRs is both timely and important.

As a lifestyle behavior, physical activity (PA) is a modifiable risk factor for CVD and leads to improved long-term health outcomes among KTRs [8]. Among the general population, PA is independently associated with a decreased risk of chronic disease morbidity and all-cause mortality, improved cardiometabolic risk factors and many other improved health outcomes [9]; conversely, insufficient PA is associated with an increased risk of the aforementioned risks and risk factors [10–13]. Among KTRs, insufficient PA (a major risk factor for CVD) appears to be highly prevalent [14, 15]. A recent review reported that PA among KTRs is lower than in the overall general population and other populations with chronic diseases [16]. For example, elderly KTRs were reported to have poorer physical functioning compared with other elderly adults with chronic diseases such as congestive heart failure and chronic obstructive pulmonary disease [17]. Generally there is a lack of studies reporting PA levels among KTRs compared with the general population. One study reported that PA was ~18–35% lower than age-matched healthy participants in the pretransplantation period. While PA levels increased posttransplantation, levels remain lower than in healthy participants [18].

PA has been found to be inversely associated with cardiovascular and all-cause mortality [19], ergo PA is associated with improved outcomes following kidney transplantation. For example, one study found that PA was associated with improved glomerular filtration rate [20], while another study found that pretransplant PA measured shortly after transplantation was associated with a reduced risk of death at follow-up [21]. A cross-sectional study with 1975 participants in Greece (Attica) identified insufficient PA posttransplantation as a risk factor for CVD events [22]. To date, only one study conducted a longitudinal analysis of PA and CVD risk among KTRs (among a Dutch cohort) [19]. The authors surveyed self-reported PA among 540 KTRs and found an increased rate of CVD mortality over a follow-up period of 5.3 years across tertiles of PA adjusted for other CVD risk factors and muscle mass. The mortality rate of the least active tertile was 24%, while the most active tertile had a mortality rate of 5.6%. However, the study lacked the inclusion of some clinical variables that can affect KTR outcomes [e.g. transplant donor type, transplant vintage (i.e. time elapsed since transplant surgery), stage of CKD].

Available evidence, while scarce, indicates that KTRs are generally insufficiently physically active. No large-scale, prospective, multinational studies have been performed to examine PA and CVD risk among KTRs. In addition, available studies typically examine end-stage renal disease patients on dialysis and neglect those who have gone on to undergo transplantation [14, 23–26]. A recent multinational cross-sectional study with 4033 KTRs reported that younger age, deceased donor source, lower pulse pressure and no history of diabetes was associated with greater amounts of PA [27]. However, the study focuses on examining the association between PA and CVD risk factors and not the development of hard, centrally adjudicated CVD outcomes. Hence this article seeks to fill these gaps in research as it directly addresses the clinical implications of how PA affects adjudicated CVD outcomes; centralized adjudication also addresses any potential global differences across study sites.

Efforts to improve PA among KTRs have been limited. To our knowledge, only one intervention study has attempted to promote PA among KTRs: a recent intervention trial to improve pedometer-based PA among KTRs found that participants in the experimental condition reported improved blood pressure and fasting glucose compared with participants in the control group, but overall adherence to PA guidelines were relatively low (58% for the experimental group and 37% for the control group, respectively) [8]. Another recent study focused on prehabilitation in the pretransplantation period and found that improved PA levels led to a shorter length of hospital stay after surgery (however, the study did not examine long-term outcomes posttransplantation) [28]. Considering that KTRs have a higher risk of CVD mortality compared with the general population [3, 4, 29], and PA has powerful benefits on health in the general population, more evidence is needed to explore the association between PA and risk of CVD mortality (and all-cause mortality). Therefore the aim of this study was to examine the relationship between PA and the development of CVD events, CVD mortality and all-cause mortality among a large, multinational, multiethnic cohort of KTRs.

MATERIALS AND METHODS

FAVORIT study overview and study participants

We examined data collected between August 2002 and January 2012 from the Folic Acid for Vascular Outcome Reduction in Transplantation (FAVORIT) Trial (National Institute of Health Clinical Trial Number U01DK061700) [30]. The FAVORIT trial was a randomized intervention trial conducted in 27 clinical sites in the USA, 2 clinical sites in Canada and 1 site in Brazil. Details of the study design have been published previously [31]. All study participants provided informed consent in accordance with the procedures of the institutional review boards at each of the study sites.

Men and women 35–75 years of age who were clinically stable recipients of a renal transplant were recruited. A review of the medical chart showing that the patient’s current graft had been functioning for ≥6 months with no clinical indication of kidney function deterioration established the study entry criterion of stable kidney function. Additional entry criteria regarding creatinine clearance and homocysteine levels have been described elsewhere [27]. The medical history and demographics (i.e. participant’s age, gender and race/ethnicity) were recorded during the screening and/or baseline visit. The National Institutes of Health Policy on Reporting Race and Ethnicity Data was used to define race and ethnicity, and participants with origins in Africa were considered to be African American [32]. Medications taken regularly during the past month were verified by review of medication bottles, patient lists or verbal self-report.

Laboratory methods

Fasting or nonfasting blood samples were collected and prepared for central analysis of total cholesterol (TC), triglycerides, high-density lipoprotein cholesterol, low-density lipoprotein (LDL) cholesterol, creatinine and glucose.

PA

PA and sedentary behavior were measured at baseline (i.e. study intake) using the Yale Physical Activity Survey (YPAS) [33]. Five activity dimensions were calculated according to the procedures described by DiPietro et al. [33]: vigorous activity, leisurely walking, moving on feet, standing and sitting. Results were then summed to calculate the Physical Activity Summary Score (possible range 0–142) [33]. A higher score indicates higher levels of PA. Data on the standing dimension were not captured prior to April 2005. A revised summary score (possible range 0–127) was computed to evaluate PA for all participants. Principal components analysis was conducted and revealed that the revised summary score explained 91.2% of the variance (‘standing’ explained 9.8% of variance) and hence we omitted ‘standing’ as part of the YPAS for the purposes of this study. The YPAS demonstrated adequate repeatability and some validity by correlating with several physiologic variables (e.g. percent body fat and maximum rate of oxygen consumption), reflecting habitual PA. We created tertiles based on data distribution and the following scores: first tertile: 0–21, second tertile: 22–38, third tertile: 39–127, where a higher tertile corresponds to more PA. The rationale for tertile grouping was due primarily to ‘clumping’ at certain score values and is described in detail elsewhere [27]. Briefly, the key PA characteristics are that (i) participants in the highest tertile of PA participated in more vigorous PA and walking compared with the lower two tertiles (the same trend is observed when comparing the middle tertile with the lowest tertile) and (ii) there were no differences across tertiles with regard to household, occupational and sedentary activities.

CVD and coronary heart disease risk factors

CVD was defined as the presence of any of the following history of diagnoses or procedures that occurred during follow-up: myocardial infarction; stroke; coronary artery bypass graft surgery; percutaneous transluminal angioplasty of one or more of the coronary, renal or lower extremity arteries; lower extremity amputation above the ankle or abdominal or thoracic aortic aneurysm repair. These clinical events were centrally adjudicated. A history of diabetes before or after transplantation by self-report or noted in the medical record was defined as positive for diabetes, and this included diabetes resolved through pancreas transplantation. We obtained participant-reported cigarette smoking status (current, former or never smoker). For current or former smokers, pack-years of smoking were determined. For the current analysis, current smoking (excluding former smokers) was considered positive as a risk factor. Dyslipidemia was defined as TC ≥5.18 mmol/L (200 mg/dL).

Sitting blood pressure was measured twice in the same arm using standard clinical procedures and the results were averaged. The participant was considered to have hypertension if the participant took one or more antihypertensive medication or if the average of the measured sitting systolic blood pressure (SBP) was ≥130 mmHg or diastolic blood pressure ≥80 mmHg.

Height and weight were measured with the participant fully clothed with shoes removed, and body mass index (BMI) was calculated and classified according to the National Heart, Lung and Blood Institute criteria: underweight ≤18.5 kg/m², healthy weight 18.5–24.9 kg/m², overweight 25.0–29.9 kg/m² and obese ≥30.0 kg/m². A BMI ≥25 kg/m² was considered positive for the coronary heart disease risk factor of overweight/obesity [34].

Statistical analysis

We examined the distribution of baseline covariates by tertiles of PA and assessed the differences using the chi-squared test for categorical variables, analysis of variance one-way test for normal continuous variables and Kruskal–Wallis test for nonnormal continuous variables. Kaplan–Meier survival curves by tertiles were constructed for each outcome (CVD event, CVD death and all-cause mortality) to compare unadjusted event rates by tertile groups, and the log-rank test was used to assess differences between tertiles. We examined linear trends over tertiles using the Cochran–Armitage trend test. Cox regression models were used to estimate the hazard ratios (HRs) of the outcomes by tertile. Both adjusted and unadjusted models were examined. We preselected covariates in the model based on background knowledge and the previous results [27]. Baseline covariates were used to adjust the HRs: age, gender, race, country of residence, SBP, renal transplant type, time since transplant, prevalent diabetes, prevalent CVD, overweight, smoking, glucose level, TC, LDL and creatinine. We tested possible violations of the proportional hazard assumption over all the models [35]. The assumption was not met for (i) time since transplant and (ii) prevalent diabetes variables, hence we set time-dependent coefficients for these two variables [36]. All statistical analyses were performed using R version 3.5.1 (R Foundation, Vienna, Austria) [37]. We address missing data via multiple imputations using chained equations in the mice package within the R software. We set the number of imputation data as 5 and the number of maximum iterations as 20. We followed the default option of the software for other procedures and we used the same Cox model and covariates as we described in the article.

RESULTS

Among 3915 eligible participants, 865 participants were excluded from analysis due to missing data. Table 1 displays the characteristics of the 3050 participant analytic sample, including the percentage of participants who developed CVD events, CVD deaths and all-cause mortality. Baseline characteristics of the analytic sample and missing cases were compared and missing participants were more likely to be from Canada and had higher glucose, TC and creatinine levels. Reasons for loss to follow-up were not recorded in our study.

Table 1.

Incidence of CVD events, deaths and all-cause mortality and baseline characteristics of FAVORIT study participants

Level	Complete	Missing	P-value
n = 3915	(N = 3050)	(n = 865)
Demographics
Gender, n (%)
Male	1903 (62.4)	562 (65.0)	0.178
Female	1147 (37.6)	303 (35.0)
Race, n (%)
White	2292 (75.1)	664 (78.5)	0.098
Black	564 (18.5)	130 (15.4)
Country, n (%)
Other	194 (6.4)	52 (6.1)
USA	2201 (72.2)	619 (71.6)	<0.001
Canada	316 (10.4)	167 (19.3)
Brazil	533 (17.5)	79 (9.1)
Donor type, n (%)
Cadaver	1767 (57.9)	482 (57.9)	0.999
Living	1283 (42.1)	351 (42.1)
Diabetes history, n (%)
No	1820 (59.7)	512 (59.3)	0.887
Yes	1230 (40.3)	351 (40.7)
CVD history, n (%)
No	2427 (79.6)	703 (82.2)	0.095
Yes	623 (20.4)	152 (17.8)
Overweight, n (%)
No	850 (27.9)	186 (24.3)	0.053
Yes	2200 (72.1)	579 (75.7)
Smoking history, n (%)
Never	1508 (49.4)	384 (46.7)	0.265
Current	331 (10.9)	102 (12.4)
Ever	1211 (39.7)	337 (40.9)
Age (years), mean (SD)	51.76 (9.43)	52.28 (9.41)	0.158
Clinical variables
Systolic BP, mean (SD)	136.21 (19.95)	135.70 (19.11)	0.516
Vintage, median (IQR)	4.07 (1.67–7.45)	3.74 (1.66–7.87)	0.79
Glucose, mean (SD)	116.13 (57.45)	123.12 (61.49)	0.002
TC, mean (SD)	182.10 (40.78)	193.48 (53.08)	<0.001
LDL, mean (SD)	100.68 (33.91)	99.32 (33.79)	0.364
Creatinine, mean (SD)	1.66 (0.56)	1.72 (0.59)	0.005
YPAS sitting score, mean (SD)	2.31 (0.98)	2.33 (0.98)	0.594
CVD events, deaths, all-cause deaths
CVD events, n (%)
No	2624 (86.0)	737 (85.2)	0.573
Yes	426 (14.0)	128 (14.8)
All-cause deaths, n (%)
No	2693 (88.3)	756 (87.4)	0.51
Yes	357 (11.7)	109 (12.6)
CVD deaths, n (%)
No	2925 (95.9)	830 (96.0)	0.999
Yes	125 (4.1)	35 (4.0)

Boldface indicates statistical significance.

IQR, interquartile range.

Participant age ranged between 38 and 72 years and about one-third were female. Less than a quarter of the participants identified as a member of a racial minority group and less than a quarter were from sites outside of the USA. More than half the sample (57.9%) received a renal transplant from a cadaver donor. The majority of the sample (72.1%) were either overweight or obese. Only about half (49.4%) of the sample were never smokers.

During follow-up (2500 days), 426 participants experienced a CVD event, 125 participants died (of CVD origin) and 357 participants died from any cause (Table 1). All the outcomes had significant P-values of the Cochran–Armitage test (i.e. CVD events 0.0074, all-cause mortality 0.0001, CVD mortality 0.0017), suggesting a linear trend between the outcome events and tertiles of PA. Figures 1–3 display the survival curves for CVD events, CVD mortality and all-cause mortality, respectively. Participants in the highest YPAS tertile (Tertile 3) had the lowest probability of developing a CVD event, CVD mortality or all-cause mortality. Adjusted models of Cox regression (Table 2) revealed that compared with the lowest tertile, the highest tertile had significantly lower risk of CVD events {HR 0.76 [95% confidence interval (CI) 0.59–0.98]}, all-cause mortality [HR 0.62 (95% CI 0.47–0.82)] and CVD mortality [HR 0.58 (95% CI 0.35–0.96)]. Compared with the middle Tertile 2 scores, the Tertile 3 of PA scores had significantly lower risk of CVD events [HR 0.76 (95% CI 0.59–0.98)], all-cause mortality [HR 0.61 (95% CI 0.46–0.81)] and CVD mortality [HR 0.50 (95% CI 0.31–0.82)]. Results were similar in unadjusted models, but the observed effect sizes were smaller.

FIGURE 1.

Kaplan–Meier survival curves of CVD events (including deaths) across three tertiles of PA in kidney transplant recipients followed over a mean of 3.7 years. Longest follow-up time is 2493 days.

Table 2.

HRs (95% CIs) of CVD outcome, all-cause deaths and CVD deaths of study participants by YPAS tertile

YPAS tertiles	HR (95% CI)
	CVD events	All-cause deaths	CVD deaths
Unadjusted
YPAS Tertile 2 versus 1	0.819 (0.656–1.021)	0.874 (0.690–1.106)	0.955 (0.647–1.412)
YPAS Tertile 3 versus 1	0.572** (0.449–0.728)	0.498** (0.379–0.655)	0.432** (0.265–0.705)
YPAS Tertile 3 versus 2	0.698** (0.546–0.893)	0.570** (0.433–0.751)	0.452** (0.279–0.734)
Fully adjusted
YPAS Tertile 2 versus 1	0.999 (0.797–1.252)	1.011 (0.793–1.288)	1.151 (0.769–1.723)
YPAS Tertile 3 versus 1	0.760* (0.592–0.976)	0.619** (0.466–0.821)	0.577* (0.349–0.955)
YPAS Tertile 3 versus 2	0.761 (0.594–0.976)	0.612 (0.464–0.809)	0.501 (0.307–0.818)

All the models adjusted for age, sex, race, country, SBP, donor type, vintage, diabetes history, CVD history, overweight, smoking history, glucose, TC, LDL and creatinine.

^*P < 0.05,

^**P < 0.01 when compared with reference tertile.

FIGURE 2.

Kaplan–Meier survival curves of CVD mortality across three tertiles of PA in kidney transplant recipients followed over a mean of 3.7 years. Longest follow-up time is 2493 days.

FIGURE 3.

Kaplan–Meier survival curves of all-cause mortality across three tertiles of PA in kidney transplant recipients followed over a mean of 3.7 years. Longest follow-up time is 2493 days.

To assist in our interpretation of the lack of significant differences between Tertiles 1 and 2, we examined the kernel density plot (data not shown) of the YPAS score by tertile, which indicates that the distribution of the YPAS score is similar between Tertiles 1 and 2, while the YPAS score distribution of Tertile 3 is distinctively different from Tertiles 1 and 2. That suggests that participants in Tertiles 1 and 2 have a comparable risk of outcomes that is driven by PA.

DISCUSSION

We report an association between PA and a reduced risk of CVD events, CVD mortality and all-cause mortality among KTRs. To our knowledge, this is the first study to report these observed associations among a large, multinational sample of KTRs. The associations were observed even when controlling for traditional CVD risk factors (age, gender, race, country of residence, SBP, renal transplant type, time since transplant, prevalent diabetes, prevalent CVD, overweight, smoking, glucose level, TC, LDL and creatinine), which suggests the importance of PA in reducing the risks of CVD and all-cause death among KTRs. In addition, overall, while our participants were relatively inactive compared with the general population, our results indicate that even light PA may be associated with lower CVD risks, in line with the 2018 Physical Activity Guidelines Advisory Committee report [38].

PA is recommended to reduce the risk of CVD independently and by improving CVD risk factors such as controlling blood pressure [39], lowering triglycerides [40] and improving glucose tolerance and insulin sensitivity [41, 42]. Conversely, insufficient PA is strongly associated with higher body fat (including among older adults) [43] and visceral fat [40], which in turn impacts insulin sensitivity and inflammatory markers. While these mechanisms were examined either among general or non-KTR populations, we may postulate that PA reduces CVD risk in the same way among KTRs. Indeed, a systematic review of observational studies of PA among KTRs reported that PA was associated with better cardiorespiratory fitness, quality of life and lower body fat [25]. Regardless, more studies (especially longitudinal studies) are needed to determine the mechanisms of CVD risk reduction among KTRs specifically.

In terms of exercise training, a Cochrane review of a handful of randomized controlled trials of exercise training in people with chronic kidney disease (including KTRs) concluded that there was evidence of enhanced exercise capacity, improved blood pressure, health quality of life and some nutritional biomarkers [44]. Studies of the effects of exercise on blood pressure and blood lipoproteins, kidney function and immune function have been equivocal [44], but more trials specifically examining the KTR population are needed. In particular, attention should be paid to the dose of exercise among KTRs, as it may be highly variable. Our study results indicate that PA benefits KTRs in terms of CVD health and survival, but, given the generally low levels of exercise in our sample, it seems that exercise training has yet to be comprehensively incorporated into routine care for KTRs after transplantation.

Study strengths include a longitudinal cohort design, allowing the identification of PA with a reduced risk of the development of hard, centrally adjudicated CVD outcomes. Limitations of our study include the self-report nature of the PA, which has been found to result in a degree of overestimation of the performance of moderate-to-vigorous PA and underestimate sedentary behavior due to social desirability bias [45, 46]. In addition, PA scores were only obtained at baseline. Hence it is unclear if the PA changes over time are associated with CVD risk. Furthermore, while the YPAS is useful for examining PA among geriatric and clinical populations, it does not permit the interpretation of PA levels in the context of national PA guidelines. In addition, our study did not measure physical functioning among the sample (which may be a cause and effect of physical inactivity), which has been found to be associated with PA in a dose-dependent manner [47].

In conclusion, the observed associations between PA and CVD indicate the importance of PA in reducing CVD risk among KTRs. Future studies could focus on (i) questionnaire and/or device measurements of PA over time in association with CVD and other health outcomes and (ii) the development of effective intervention trials to improve PA among KTRs, with special attention in examining their effectiveness in altering CVD biomarkers, as well as identifying factors associated with adherence. In addition, such efforts should address the challenges of developing PA programs for KTRs due to wide-ranging KTR comorbidities, psychosocial and socioeconomic factors and long-term use of immunosuppression that in turn affects the recommended frequency, duration and intensity of PA [16]. While beyond the scope of our present discussion, intervention designs with KTRs should also consider targeting psychosocial barriers (e.g. quality of life) and other lifestyle behaviors, including diet [48], as part of weight management strategies to prevent CVD.

FUNDING

This study was funded by a National Institute of Health grant (NIH U01DK061700). A.W.K. receives funding as a predoctoral fellow from the American Heart Association (18PRE34060015).

AUTHORS’ CONTRIBUTIONS

A.W.K. led the manuscript preparation, formulated the research hypothesis and was the secondary statistical analyst. A.G.B. was the principal investigator of FAVORIT, led the original data collection and developed the rationale for the study. H.K. served as the principal statistical analyst. C.B.E. served as the consultant and subject-matter expert on the topic of PA and cardiovascular health. R.G. was the site principal investigator at Rhode Island Hospital and together with A.G.B. piloted all hypotheses in the FAVORIT trial. J.W.K. served as the consultant and subject-matter expert on the context of PA among KTRs. M.A.P. led the process of adjudicating the CVD events. P.M.R. served as a consultant on the analysis approach and literature review. C.E.G. led the process of conceptualizing this article and formulating the research hypothesis and oversaw the manuscript preparation process. All authors provided input on the manuscript outline, drafted sections of the manuscript and revised multiple drafts.

CONFLICT OF INTEREST STATEMENT

This work has not been published previously in whole or in part in any other journal. M.A.P. receives research support from Novartis and serves as a consultant for AstraZeneca, Corvidia, DalCor, GlaxoSmithKline, Jazz, MyoKardia, Novartis, Novo Nordisk, Pfizer, Roche, Sanofi, Servier and Takeda and has equity in DalCor. No other authors have any conflict of interest to disclose.