Peripheral vascular disease (PVD) is a strong predictor of morbidity and mortality in dialysis patients,1-3 but only a few studies have focused on the impact of this entity on survival in kidney transplant (KT) candidates.4-9 In addition, the consequences of PVD on waitlist mortality in the presence of clinical conditions compiled in the Charlson comorbidity index (CCI) plus other uremia-related comorbidities and community risk factors have not been fully studied.1,4,7,10,11
When assessing mortality in waitlist patients for KT, interpretation of results of survival analyses may be misleading since transplantation may be considered as a competing risk. Indeed, KT may hinder or modify the chance the event of interest occurs (eg, death). Thus, conventional methods for survival analysis, ignoring the competing events (eg, conventional Cox regression analysis), could be inappropriate, leading to overestimating the risk of death, especially in the presence of competing risks, as would be expected in KT candidates. In these cases, alternative methods specifically designed for analyzing competing risks data should then be applied,12 but there is a lack of information on using competing risk models to estimate survival analyses in this particular population.
The aim of this cohort study in KT candidates was to assess the association between survival and PVD in the presence of other comorbidities (evaluated with the CCI), plus uremia-related comorbidities and certain community risk factors, by using conventional methods for survival analyses and competing risk analyses.
MATERIALS AND METHODS
This longitudinal cohort study involved 22 497 renal patients included in the Information System of the Andalusian Transplant Coordination (SICATA) from the Andalusian Registry of Renal Patients between January 1, 1984, and July 31, 2012, because of initial onset of dialysis. Data from all centers in Andalusia, a region located in the south of Spain, were entered into SICATA with follow up information through July 31, 2012. The database is updated annually and the degree of compliance of data concerning waitlist patients, including follow-up, was almost universal. Baseline clinical and demographic data of the study population have been previously reported.10 Briefly, we excluded 18 560 patients not listed for KT at entry to dialysis, including preemptive transplantation, and 86 patients younger than 18 years. We therefore assessed a final population of 3851 adult KT candidates, who either received a KT (n = 1975) and were therefore censored, or were on the waitlist (n = 1876) at any time during the study period (Figure 1). The overall median time on the waiting list was 21.2 months (interquartile range, 11-37.4). The clinical characteristics of waitlisted patients and KT recipients are shown in supplementary data (Table S1, SDC,http://links.lww.com/TP/B296). Inactive status was defined as candidates not suitable for KT, for whatever reason, at the time of waitlist inclusion or while on the waiting list (incomplete workup, medical noncompliance, inappropriate weight, temporarily too sick, and so on), but still an appropriate patient to remain on the transplantation waiting list. Consequently, we also took into account “inactive status” within the waitlist (n = 316) when performing survival analyses because these patients had not been definitively removed from the waitlist. Thus, we evaluated survival of all patients who remained on the waitlist during follow-up, including inactive status patients.
Recommendations for keeping a candidate on the waitlist or removing someone from it in our center are broad rather than specific, and they include assessment of age, life expectancy, psychosocial factors, and comorbidities, according to clinical practice guidelines on waitlisting for KT.13,14 For instance, conditions that may make a transplantation inadvisable include factors related to older age, short life expectancy, unsolved psychosocial disorders, severe cardiovascular diseases, or advanced malignancies.
The following data were recorded at the beginning of dialysis: (i) demographic and clinical data: age, gender, cause of renal disease (categorized as diabetes, glomerulonephritis, polycystic disease, nephrosclerosis, interstitial nephritis, unknown, or other causes), and comorbidities included in the CCI15: myocardial infarction and congestive heart failure (grouped as cardiac disease), hemiplegia, diabetes, connective tissue disease, mild liver disease (considered as the presence of hepatitis B surface antigen or positive hepatitis C virus antibodies without cirrhosis), cirrhosis, chronic pulmonary disease, peptic ulcer, any tumor without metastasis and HIV infection; (ii) conditions and comorbidities inherent to the uremic status: presence of a central venous catheter (CVC) at dialysis entry regardless of dialysis type, dialysis modality at entry (hemodialysis versus peritoneal dialysis), time on waitlist (defined as time from inclusion on list until end of follow up), time on dialysis (defined as time from starting dialysis until end of follow up—including time on waitlist), and previous transplant; and (iii) community risk factors, such as early or late referral to the nephrologist (considered as a cut-off time of 6 months) and employment status (employed vs. unemployed or retired because of age or disability). Finally, the waitlisted year was also registered, grouped in 4- or 5-year periods (1984-1987 vs 1988-1992 vs 1993-1997 vs 1998-2002, 2003-2007 vs 2008-2012).
Identification of PVD
Patients with clinical evidence of symptomatic PVD at waitlist entry were identified by searching SICATA using the following diagnostic categories: amputation, peripheral aneurysm, arterial embolism and thrombosis, peripheral angioplasty, atherectomy, or endarterectomy and peripheral bypass, according to diagnosis code from the International Classification of Diseases, Ninth Revision, Clinical Modification.16,17 Thus, patients were classified as having PVD if they had a previous diagnosis of PVD or a known prior amputation attributable to PVD. Data about PVD compiled in SICATA were provided by each individual center in Andalusia.
Medical record review was performed according to Spanish law, approved by the Ethics Committee of Carlos Haya Regional University Hospital and conducted according to the Declaration of Helsinki.
Patient death after placement on the waitlist was the clinical endpoint. Survival was measured in months from the date of waitlist inclusion with patients censored at the time of permanent or temporal waitlist withdrawal for any reason, including KT, inactive status or last follow up (December 31, 2012). We took into account competing events in our study. Thus, survival analyses were performed using both standard Cox regression models (ignoring competing events) and a competing risk approach, where KT and inactive status condition were treated as competing events. Survival data were available for the whole population, and all deaths were recorded accurately. Survival analysis considered the whole study population; that is, all active and inactive status patients on the waitlist. Survival was also analyzed excluding inactive status and diabetic patients.
Continuous data were summarized as mean ± standard deviation or median and interquartile range. Comparisons of continuous variables between patients with and without PVD were performed using unpaired t test. Absolute and relative frequencies were used to describe categorical variables, which were compared between both groups using the χ2 test. The cumulative incidence of death over time according to the presence or absence of PVD was displayed by using Nelson-Aalen estimates with their 95% confidence intervals (95% CI).18 We used a Cox proportional hazard model to analyze patient survival and a competing risk approach as sensitivity analyses where KT and inactive status condition were treated as competing events. Additionally, we performed competing risk models for cardiovascular mortality by grouping KT, inactive status, and other causes of death as competing events. Explanatory variables included comorbidities derived from the CCI, age, sex, presence of a CVC regardless of dialysis modality, time on dialysis, late referral to nephrologist, employment status, dialysis modality, previous transplant and listing periods. Given the strong association between diabetes and mortality in renal patients, and that more than half of patients with PVD had diabetes, we entered the “diabetes” condition (vs nondiabetes) as a covariate in the multivariate survival analyses. Proportionality assumption in the model was assessed by visual inspection of the log-log survival plots. Covariates included in the model did not violate the proportionality assumption. The interaction between PVD and age, diabetes, time on dialysis, and cardiac disease was also considered for all cause-mortality.
We repeated the same analysis using competing risk regression because KT and inactive status are events that compete with our clinical end-point. In this model, KT and inactive status condition were treated as a competing risk and study end as an administrative censoring. This approach provides better performance than standard Cox regression models for survival analysis assessment when competing risks are present,12,19 as would be expected in waitlist patients. In particular, we used multivariable regression modelling that was applied directly to the cumulative incidence function for particular use in competing risks analyses, as described by Fine and Gray.19 This model was fitted using the stcrreg Stata command. These analyses were also performed excluding diabetic patients.
All analyses were performed using SPSS version 15.0 (SPSS, Inc., Chicago, IL) and Stata version 13.1 (StataCorp LP., College Station, TX). A P value less than 0.05 was considered significant.
Some of the clinical data were not available in all patients. In general, there were relatively few missing values. Missing data were treated by means of random effects simple imputation; that is, to produce just 1 imputed value for each missing item at random.20 We performed a separate analysis for the subset of patients with no missing data (93% of the population) in covariates included in the final survival analyses to confirm that missing values did not have major effects on the results.
Of the 3851 patients included in the study, 308 (8%) had a diagnosis of PVD at the start of dialysis. The prevalence of PVD among nondiabetic patients was 4.5%, whereas among patients with diabetes it was 25.3% (P < 0.0001). Table 1 summarizes baseline clinical and demographic characteristics between patients with and without PVD in the entire population. A similar proportion of patients receiving hemodialysis at entry was observed between both groups. Patients with PVD were older and more likely to have diabetes, chronic pulmonary disease, peptic ulcer, and other cardiovascular disorders, such as cardiac disease and hemiplegia. Consequently, a higher CCI score was observed in the PVD group compared to patients without PVD. In addition, a greater proportion of males, CVC and unemployed status were also documented in PVD patients. As expected, a higher proportion of waitlist patients without PVD at the beginning of dialysis received a KT (53%), whereas only 30% of patients with PVD at the beginning of dialysis received a KT (Table 1). Figure 2 displays the cumulative incidence of KT in patients with and without PVD.
The overall median follow-up time was 22 months (interquartile range, 12-48 months). No differences were observed in time on dialysis between patients with and without PVD (Table 1). As shown in Table 2, of the 1876 patients on the waiting list, 446 (24%) died due, mainly, to cardiovascular diseases (25.3%), followed by infectious complications (19.3%), neoplasm (7.2%), and hepatic disorders (3%). Deaths of uncertain origin and miscellaneous causes accounted for 29% and 17.2%, respectively. Of note, patients with PVD, showed a higher all-cause mortality rate than patients without PVD (45 vs. 21%; P < 0.0001). Finally, a trend toward a higher cardiovascular mortality was seen in patients with PVD compared to patients without PVD (Table 2).
Among patients on the waiting list (n = 1876) who died (n = 446; 24%), 272 (61%) died within 2 years after listing, with the rate decreasing in the following years (χ2 = 10.4, P = 0.001). Indeed, the median time on the waitlist of patients who died was significantly lower than patients who remained alive on the waitlist and those who received a KT (Table S2, SDC,http://links.lww.com/TP/B296). Additionally, patients who died were older and had a higher burden of comorbidities than the rest. Likewise, among patients with PVD who died (n = 96), a similar proportion of deaths (66%) occurred during the first 24 months after placement on the list.
Figure 3A shows the Nelson-Aalen plot describing the cumulative hazard of deaths over time in waitlisted patients with and without PVD by the slope of the cumulative curve. For instance, the cumulative incidence of all-cause mortality at 2 years in patients with and without PVD was 23% and 6.4%, respectively (P < 0.001). Similarly, marked differences in the mortality were also observed between both groups when excluding diabetic patients (Figure 3B). Of note, when assessing only waitlist diabetic patients at the beginning of dialysis, a significantly decreased survival was also observed in patients with PVD compared with those without PVD (Figure 3C).
The conditions associated with mortality in univariate analysis (all variables studied except for gender) were included in the multivariate survival analyses. Table 3 displays adjusted patient survival models using Cox proportional hazards regression and competing-risk models. The presence of PVD at the start of dialysis was associated with a higher risk of death in both the conventional Cox regression and when KT and inactive status condition were treated as competing events after adjusting for age, comorbidities included in the CCI, diabetes, uremia-related factors, unemployed status, late referral and listing period (Table 3). The presence of PVD at waitlist entry resulted in a 1.9-fold increased risk of all-cause mortality during follow up. In addition, PVD was associated with cardiovascular mortality after adjusting for confounders using a competing risk approach (subhazard ratio [SHR], 2.9; 95% CI, 1.8-4.9). There was no significant effect of listing periods on survival in our multivariate models. No significant interaction was found between PVD and diabetes in our final regression analyses (P = 0.332), and similar results were seen after excluding both inactive waitlisted and diabetic patients. Likewise, we did not find an interaction between PVD and age (P = 0.125) or time on dialysis (P = 0.219). However, we found an interaction between PVD and cardiac disease (P = 0.045); that is, the association between PVD and mortality was stronger in waitlisted patients without cardiac disease (SHR, 2.2; 95% CI, 1.6-3.1, P < 0.001) versus those with cardiac disorders (SHR, 1.5; 95% CI, 0.9-2.5; P = 0.120) (Figures 4A-B).
In this large cohort study of waitlist candidates for KT, we analyzed for the first time the negative impact on survival of PVD in the presence of other comorbidities included in the CCI, plus uremia-related factors and community risk factors, by using standard methods for survival analyses and a competing risk approach. We obtained comprehensive clinical information from the Andalusian Registry, where clinical data are accurately collected. Thus, continued efforts to identify PVD at entry to waitlist could help optimize allocation in high-risk candidates for KT and guide appropriate treatments of this condition for minimizing mortality in these patients.
Although PVD has been associated with a higher mortality in the dialysis population,1,2 the relationship between PVD and death in waitlisted patients for KT has not been extensively studied. Among the KT candidates of our study, a diagnosis of PVD was associated with an almost 2-fold increase in the risk for all-cause death. Additionally, PVD was predictive of cardiovascular mortality. To account for confounding on mortality, we adjusted survival analyses for baseline comorbidities which have been found to be associated with waitlist mortality in previous studies.4,7,10 Since the prevalence of comorbidities in incident dialysis patients has increased in recent years,21,22 adjustments for listing periods were also performed. Our results are consistent with previous studies in both waitlist patients and KT recipients in other regions,4,8,23 but we also included important uremia-related conditions, such as CVC, previous transplant or dialysis technique at entry, as well as community risk factors that have been associated with a higher risk of death in dialysis patients.11,24-27 Indeed, except for dialysis technique, waitlist patients with a CVC also showed a higher risk of death in our study. In addition, unemployment status and late referral to the nephrologist were associated with mortality in our KT candidates by using a competing risk approach, which suggests that patients likely had a marginal health status or differences in quality and access to healthcare, particularly in the predialysis period, as expected in an older and fragile population.28 Finally, a history of a previous transplant was also significantly associated with a higher mortality. An association between high mortality risk and failed transplant has been reported among patients on dialysis therapy.27,29 Taken together, whether a higher number of comorbidities in older patients, including uremia-related factors and community risk factors, could have contributed to more severity of PVD in our study is undetermined.
We used a competing risk analysis, which may be more appropriate for analyzing waitlist patient survival in the presence of competing events,12,30 as would be expected in KT candidates. Indeed, patients who experience a competing event, such as KT or removal from the list for any reason, remain in the risk set (instead of being censored), although they are no longer at risk of the event. In a competing risks setting, this approach shows better performance than the standard Cox proportional hazard model when assessing survival in epidemiological studies. Thus, failure to account for such competing events results in an overestimate of the hazard.31
The prevalence of PVD in our cohort of waitlist patients (both nondiabetic and diabetic) was similar to that reported previously for this particular population.4 However, the prevalence rates found in our waitlist patients are lower than those reported for the larger hemodialysis population,1,2 which is not surprising given that patients who are on the waitlist for KT are, generally, selected to be younger and healthier.
All-cause and cardiovascular mortalities were increased in our patients with PVD. This is consistent with previous reports in the general population and renal patients.7,23,32,33 Peripheral vascular disease may reflect progression of atherosclerotic disease, especially in an older population with a higher burden of diabetes and cardiovascular disorders, as seen in our patients with PVD. Consequently, among patients who died, almost two thirds of deaths occurred within 2 years after listing due, mainly, to cardiovascular diseases, as reported.10 Accordingly, patients who died were older and had a higher burden of comorbidities than the rest. In previous reports nearly 50% of older waitlist patients (>50 years) died prematurely within the first years after listing before receiving a KT.7,9,34,35 Another plausible explanation for a high mortality in our waitlisted patients is the long study period where information on mortality data extends back to the early 80s.
Finally, we cannot rule out that other inflammatory markers (eg, high C-reactive protein and serum albumin levels) or mineral metabolism disorders (eg, high serum phosphorus levels) could have contributed to the causal pathways for increased all-cause mortality, as reported.3 This hypothesis, thus, deserves further clinical research.
Although diabetes is a known risk factor for PVD in the general population and renal patients,1,32 the effect of PVD on mortality in our waitlist patients was not restricted to diabetic patients. We did not find an interaction between diabetes and PVD in our final multivariate analyses. Additionally, the cumulative incidence of all-cause mortality was significantly higher in patients with PVD compared to those without PVD, regardless of diabetic status. Likewise, the effect of PVD on mortality was more evident in waitlisted patients without cardiac disease. This suggests that PVD is a marker of cardiovascular disease that could reflect the time-dependent action of a host of candidate uremia-related proatherogenic factors affecting predominantly large blood vessels, which could be more relevant in the older population, even in the absence of diabetes or cardiac disease. In support of this argument, previous reports have documented a higher incidence of PVD in uremic patients compared with individuals without renal insufficiency.36 The fact that our waitlist patients with PVD were older and had a higher burden of cardiovascular disease than patients without PVD reinforces this view. Whether waitlist patients with PVD should proceed to more rigorous cardiovascular system assessment (eg, coronary angiography), at least in high-risk patients, remains to be determined.
This cohort study has certain limitations that are inherent to registry analyses. First, we did not record important risk factors for survival, such as blood pressure and lipid profile. Additionally, we did not collect other community risk factors, such as obesity or smoking, which have been associated with mortality in the general population and waitlist patients.11 Further research is needed to determine the impact of these important risk factors on mortality in waitlist patients with and without PVD. However, we obtained comprehensive information about comorbidities included in the CCI. Given the mandatory character of the Andalusian Registry, we collected complete information on other uremia-related clinical conditions and certain community risk factors, which provide acceptable accuracy for this analysis. Second, PVD was not defined with formal diagnostic testing. Likewise, we had no information on severity of PVD, which could underestimate the prevalence of PVD in our KT candidates, leading to potential bias in the interpretation of the results. However, given that for waitlist patients and KT recipients we used the same criteria for identifying PVD, the bias should affect all groups, preserving the validity of our results in terms of survival. Third, we included patients from a single region in the south of Europe (Andalusia, Spain), almost exclusively white, and a greater severity of PVD has been described in both African-American and Native-American uremic patients.1,37 Thus, validation of our findings is required in other centers worldwide. Lastly, in consonance with the performance of retrospective studies, we are only able to assess associations. Thus, causal inferences derived from our study should be made with caution.
In conclusion, during initial clinical assessment of candidates for KT, the presence of PVD is an independent risk factor for death in those who remain on the waitlist. Thus, accurate identification of PVD at entry to dialysis in addition to other comorbid conditions could be useful in daily clinical practice in order to encourage the implementation of targeted therapeutic strategies of risk factors and clinical decision making to prioritize high-risk waitlist patients for an age-matched organ from a deceased donor.
The authors thank the Transplant Coordination Center and all participants in each center for their assistance in data collection. The authors also thank Carmen Vozmediano and Armando Torres (RedInRen, RD12/0021/0008) for reviewing the manuscript and Ian Johnstone for linguistic assistance in the preparation of the text.
1. Jaar BG, Astor BC, Berns JS, et al. Predictors of amputation and survival following lower extremity revascularization in hemodialysis patients. Kidney Int
2. Cohen LM, Ruthazer R, Moss AH, et al. Predicting six-month mortality for patients who are on maintenance hemodialysis. Clin J Am Soc Nephrol
3. O'Hare AM, Hsu CY, Bacchetti P, et al. Peripheral vascular disease risk factors among patients undergoing hemodialysis. J Am Soc Nephrol
4. Snyder JJ, Kasiske BL, Maclean R. Peripheral arterial disease and renal transplantation. J Am Soc Nephrol
5. Grace NL, Wong LL. Should peripheral vascular disease- or diabetes-related amputation contraindicate renal transplant? Hawaii J Med Public Health
6. Akolekar D, Forsythe JL, Oniscu GC. Impact of patient characteristics and comorbidity profile on activation of patients on the kidney transplantation waiting list. Transplant Proc
7. Dusseux E, Albano L, Fafin C, et al. A simple clinical tool to inform the decision-making process to refer elderly incident dialysis patients for kidney transplant evaluation. Kidney Int
8. van Walraven C, Austin PC, Knoll G. Predicting potential survival benefit of renal transplantation in patients with chronic kidney disease. CMAJ
9. Gill JS, Ma I, Landsberg D, et al. Cardiovascular events and investigation in patients who are awaiting cadaveric kidney transplantation. J Am Soc Nephrol
10. Hernandez D, de la Nuez PC, Muriel A, et al. Clinical assessment of mortality risk in renal transplant candidates in Spain. Transplantation
11. Schold JD, Heaphy EL, Buccini LD, et al. Prominent impact of community risk factors on kidney transplant candidate processes and outcomes. Am J Transplant
12. Noordzij M, Leffondre K, van Stralen KJ, et al. When do we need competing risks methods for survival analysis in nephrology? Nephrol Dial Transplant
13. Batabyal P, Chapman JR, Wong G, et al. Clinical practice guidelines on wait-listing for kidney transplantation: consistent and equitable? Transplantation
14. Abramowicz D, Cochat P, Claas FH, et al. European Renal Best Practice Guideline on kidney donor and recipient evaluation and perioperative care. Nephrol Dial Transplant
15. Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis
16. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol
17. Romano PS, Roos LL, Jollis JG. Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives. J Clin Epidemiol
18. van Geloven N, Martin I, Damman P, et al. Estimation of a decreasing hazard of patients with acute coronary syndrome. Stat Med
19. Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc
20. Donders AR, van der Heijden GJ, Stijnen T, et al. Review: a gentle introduction to imputation of missing values. J Clin Epidemiol
21. Ng YY, Hung YN, Wu SC, et al. Progression in comorbidity before hemodialysis initiation is a valuable predictor of survival in incident patients. Nephrol Dial Transplant
22. Couchoud CG, Beuscart JR, Aldigier JC, et al. Development of a risk stratification algorithm to improve patient-centered care and decision making for incident elderly patients with end-stage renal disease. Kidney Int
23. Brar A, Jindal RM, Elster EA, et al. Effect of peripheral vascular disease on kidney allograft outcomes: a study of U.S. Renal data system. Transplantation
24. Lorenzo V, Martn M, Rufino M, et al. Predialysis nephrologic care and a functioning arteriovenous fistula at entry are associated with better survival in incident hemodialysis patients: an observational cohort study. Am J Kidney Dis
25. Winkelmayer WC, Glynn RJ, Levin R, et al. Late nephrologist referral and access to renal transplantation. Transplantation
26. Bayat S, Frimat L, Thilly N, et al. Medical and non-medical determinants of access to renal transplant waiting list in a French community-based network of care. Nephrol Dial Transplant
27. Perl J, Zhang J, Gillespie B, et al. Reduced survival and quality of life following return to dialysis after transplant failure: the Dialysis Outcomes and Practice Patterns Study. Nephrol Dial Transplant
28. Kucirka LM, Grams ME, Lessler J, et al. Association of race and age with survival among patients undergoing dialysis. JAMA
29. Rao PS, Schaubel DE, Jia X, et al. Survival on dialysis post-kidney transplant failure: results from the Scientific Registry of Transplant Recipients. Am J Kidney Dis
30. Satagopan JM, Ben-Porat L, Berwick M, et al. A note on competing risks in survival data analysis. Br J Cancer
31. Wolbers M, Koller MT, Witteman JC, et al. Prognostic models with competing risks: methods and application to coronary risk prediction. Epidemiology
32. Criqui MH, Langer RD, Fronek A, et al. Mortality over a period of 10 years in patients with peripheral arterial disease. N Engl J Med
33. Rajagopalan S, Dellegrottaglie S, Furniss AL, et al. Peripheral arterial disease in patients with end-stage renal disease: observations from the Dialysis Outcomes and Practice Patterns Study (DOPPS). Circulation
34. Delmonico FL, McBride MA. Analysis of the wait list and deaths among candidates waiting for a kidney transplant. Transplantation
35. Schold J, Srinivas TR, Sehgal AR, et al. Half of kidney transplant candidates who are older than 60 years now placed on the waiting list will die before receiving a deceased-donor transplant. Clin J Am Soc Nephrol
36. O'Hare AM, Vittinghoff E, Hsia J, et al. Renal insufficiency and the risk of lower extremity peripheral arterial disease: results from the Heart and Estrogen/Progestin Replacement Study (HERS). J Am Soc Nephrol
37. O'Hare AM, Bacchetti P, Segal M, et al. Factors associated with future amputation among patients undergoing hemodialysis: results from the Dialysis Morbidity and Mortality Study Waves 3 and 4. Am J Kidney Dis