Excellent predictors (AUC 0.80-0.89)
N-terminal pro-brain natriuretic peptide (NT-proBNP) and brain natriuretic peptide (BNP) were found to be excellent predictors of mortality in dialysis patients[22,24]. In the general population, numerous factors such as age, female sex, and obesity can influence NT-proBNP level separate from heart failure. BNP is eliminated by three main mechanisms: enzymatic metabolism, clearance through receptors, and urinary excretion. Both BNP and NT-proBNP can be elevated in dialysis patients due to poor renal clearance, concomitant heart disease, heart failure, and volume expansion[38,39]. The mechanisms underlying the link between elevated BNP and NT-proBNP with increased mortality have not been fully established. One study suggested that chronic HD induces recurrent transient episodes of myocardial ischemia that result in myocardial stunning, hibernation, and remodeling. Burton et al described in this study that the decline in cardiac function and cardiac stunning were correlated with increasing BNP levels. These myocardial changes likely have a causative pathophysiologic role in the high cardiovascular complication rates and deaths in HD patients.
Soluble urokinase plasminogen activator receptor (suPAR) and C-reactive protein (CRP) were also excellent predictors for mortality in dialysis patients[29,36]. SuPAR is a proteolytic cleavage product of the glycosyl-phosphatidylinositol anchor of the urokinase plasminogen activator receptor expressed on immune and endothelial cells. The concentration of suPAR is low in healthy individuals, but it becomes elevated in the presence of infection or inflammatory diseases. Observational studies of dialysis patients have shown that higher suPAR levels are detected in patients that have severe angiographic coronary artery disease[29,42] or after cardiovascular events, both of which are associated with significantly increased cardiovascular death. Increased suPAR levels are also associated with end-organ fibrosis, such as cirrhosis[44,45] and focal segmental glomerulosclerosis. Similarly, CRP is a nonspecific acute phase reactant that can be elevated in the setting of infection or inflammatory processes. Higher CRP levels are associated with pulmonary hypertension, which could represent crosstalk between inflammatory cells and pulmonary wall components[36,48,49]. Moreover, a population-based study demonstrated that CRP is independently associated with elevated parathyroid hormone levels (PTH) and increased mortality. Elevated PTH in CKD patients can lead to renal osteodystrophy and accelerated vascular calcification. The associations between suPAR, CRP, and increased death suggest that these markers show promise for use as standard CKD mortality predictors in the future.
Echocardiography has been shown as a useful tool in the assessment of mortality risk in CKD patients. Gan et al have demonstrated that impaired left atrial reservoir strain (LASr) is the best predictor of adverse cardiovascular outcomes in CKD stage 3-4 patients. In this assessment, the left atrium is divided into a total of 12 segments: six segments each from the apical four- and two-chamber views. LASr is measured from the average peak systolic strain from the 12 segments. Impaired LASr is defined as ≤ 23%, based on a previously reported lower limit of reference value. A few hypotheses have been postulated to link impaired LASr and mortality in the CKD population. First, the left atrium may be more predictive for mortality because it is a thin-walled chamber and may exhibit earlier alterations before the left ventricle. Second, alterations of LASr irrespective of left ventricular (LV) function have been noted in heart failure with preserved ejection fraction, further highlighting it as a more sensitive marker. The underlying mechanisms for reduced LASr are believed to be a consequence of atrial fibrosis, which could be accelerated in the setting of CKD due to the attendant high systemic inflammatory state[53,54]. These potential pathophysiologic mechanisms may explain the significance of reduced LASr as an early indicator for poor outcomes in non-dialysis CKD patients.
Rroji et al added that systolic pulmonary artery pressure (PAP) is predictive for mortality in dialysis patients. Other studies have concordantly shown that pulmonary hypertension (PH) confers a significant mortality risk in dialysis patients[55,56]. The prevalence of PH increases with CKD severity, and several studies have proposed mechanisms for the association between PH and CKD for both non-dialysis and dialysis patients. Arteriovenous fistulas for dialysis access and exposure to bio-incompatible dialysis membranes are a few of these factors that could increase the risk for PH in CKD patients[48,57]. The latter causes neutrophil activation that subsequently migrates to the lungs, resulting in increased pulmonary inflammation and vascular resistance. For non-dialysis CKD patients, CKD itself can pathophysiologically directly incite pulmonary circulatory dysfunction and remodeling through uremic toxins, systemic inflammation, altered vasoregulation and endothelial dysfunction. Altogether, prompt detection of PH via measurement of the systolic PAP could result in earlier intervention in order to prevent excess mortality in CKD patients.
Augmentation index (AIx) was predictive of mortality in CKD patients. AIx was derived from the arterial pulse wave analysis using a tonometry-based SphygmoCor device. Augmentation pressure (AP) is defined as the maximum systolic pressure subtracted from the inflection point pressure (the merging of incident and the reflected wave). AIx is defined as AP divided by pulse pressure (PP), presented as a percentage. This index represents arterial stiffness, which could be a result of various pathophysiologic mechanisms. For example, fibrosis, disruption of elastin fibers, calcification, or diffusion of macromolecules within arterial walls are potential contributors. Although AIx is useful in predicting mortality in CKD patients, given the invasive nature of this procedure, the clinical use of this parameter to monitor and prognosticate outcomes in the ambulatory setting is limited.
Acceptable predictors (AUC 0.70-0.79)
Non-dialysis population: Several predictive factors for mortality have been identified among non-dialysis CKD participants. Urine protein creatinine ratio (UPCR), eGFR, mortality risk score and malnutrition inflammation score (MIS) have been demonstrated to be predictive of mortality in this patient population[28,30,31]. It is thus not surprising that Wang et al reported that UPCR and eGFR are predictive of mortality among CKD patients. The presence of proteinuria in CKD patients has been widely associated with an increased risk of CKD progression and death. Similarly, eGFR decline represents progression of CKD severity, which independently confers a significantly increased mortality risk[59,60]. Renal interstitial fibrosis/scarring and tubular atrophy are consistent with GFR decline and degree of proteinuria. Along with these changes, the tubular epithelial cells are stimulated to synthesize reactive oxygen species, which further attracts inflammatory cells and interstitial myofibroblasts, leading to more fibrosis and scarring. These interstitial changes are associated with CKD progression, which could indirectly increase mortality. Thus, early interventions aiming to prevent proteinuria and CKD progression are essential for reducing mortality in CKD patients.
In 2019, Danial et al developed a mortality risk score for CKD patients. In this study, the authors investigated mortality from adverse drug reactions (ADRs) that specifically occurred in CKD patients. The mortality risk prediction model included the following variables: (1) History of heart disease; (2) Dyslipidemia; (3) Electrolyte imbalance; (4) Psychotic agents; (5) Creatine kinase; (6) Total number of medications; and (7) Conservative management (how the treatment was provided to subjects with ADR). The authors noted that these clinical factors are easily obtained in clinical practice, and this score allowed preemptive identification of CKD patients at risk of ADR-associated mortality. These patients may thus have a significant benefit with early clinical intervention and medication adjustment. However, this model has not been proven to be predictive of mortality outside the scope of ADR, and consequently, its applicability in general CKD patients may be limited.
MIS is another mortality risk predictor tool for both pre-dialysis and dialysis-dependent CKD patients. The MIS consists of ten elements obtained from the past medical history, physical examination, body mass index, and laboratory indices. In 2001, Kalantar-Zadeh et al demonstrated that the MIS correlated with morbidity and mortality in maintenance HD patients. Later, Jagadeswaran et al applied MIS to a pre-dialysis Indian population and found that MIS had a 56.5% sensitivity and 81% specificity (AUC 0.709; 95%CI: 0.604-0.815) for mortality prediction during a 36-mo follow-up. The mortality risk increased by 13.7% for each additional point in the MIS. These studies demonstrate the importance of malnutrition prevention for both pre-dialysis and dialysis-dependent CKD patients.
Echocardiographic factors: The ratios of early inflow velocity/annular diastolic velocity (E/E’) through the mitral valve and the LV mass index are predictive of mortality in CKD patients, regardless of dialysis status[28,36]. Mitral E/E’ ratio may be a marker of LV diastolic dysfunction or a predictor of pulmonary hypertension. Diastolic dysfunction has been associated with increased mortality in various CKD and non-CKD patient populations[63-65]. Similarly, an increased LV mass index represents LV hypertrophy, a well-established predictor for heart failure and death in CKD patients. The 2013 European Society of Hypertension/European Society of Cardiology guidelines recommend performing echocardiography to refine cardiovascular risk assessment in hypertensive patients. However, there are no recommendations specifically for CKD patients, despite studies showing an increased prevalence of LV hypertrophy with advancing CKD. Our study further adds to the available evidence supporting echocardiography for mortality risk stratification in CKD patients. It should be highlighted that a combination of echocardiographic findings should be utilized to predict mortality in CKD patients, rather than isolated findings. This concept is supported by Tripepi et al where the measurement of LV mass index in isolation does not provide significant prognostic value for CKD patients.
Real-time myocardial contrast echocardiography (MCE) is a non-invasive tool for assessment of myocardial ischemia in HD patients. During this procedure, intravenous contrast is utilized to assess myocardial perfusion adequacy, which is defined as homogenous enhancement in > 50% of myocardial wall thickness for each segment. One study showed that MCE and coronary angiography resulted in equal accuracy for anticipating combined cardiovascular endpoints and death. The authors concluded that MCE is a safe and uncomplicated test that can aid in the selection of candidates for coronary revascularization. However, this data was limited to only HD patients. The utility of MCE should be confirmed in large multi-center studies of non-dialysis CKD patients.
Comorbidities: Comorbidities may serve as significant predictors for mortality in dialysis patients. Miskulin et al investigated the index of coexisting disease (ICED) as an accurate mortality predictor in dialysis patients. In this study, ICED had greater discriminatory ability than other instruments. ICED is comprised of 19 medical conditions and 11 physical impairments; these are classified into four and three severity levels, respectively. Another unique feature of ICED is the assessment of physical limitations, which has been shown as a strong prognostic factor in dialysis patients.
The Rennes comorbidity score and modified Charlson Comorbidity Index (CCI) have also been demonstrated to be predictive for mortality among dialysis patients. The Rennes comorbidity score consists of seven components: Age, albumin level, cardiac, respiratory, and hepatic diseases, active malignancy, and walking disability. Similarly, the modified CCI consists of 16 matrices composed of a wide range of medical comorbidities found in the original CCI plus patient’s age. Pladys et al found that the original CCI (without including patient’s age) had a low ability to predict 1-year mortality. Surprisingly, the Rennes score resulted in a very impressive ability to predict 1-year mortality among dialysis patients despite its inclusion of only seven variables. Altogether, we conclude that specific comorbidities are crucial in determining mortality risk among dialysis patients. Not unexpectantly, accuracy is improved by the consideration of several high-yield comorbidities rather than relying solely on one particular comorbidity.
Miscellaneous: Acute physiologic assessment and chronic health evaluation (APACHE) III is an acceptable predictor for hospital, and 30-d mortality among HD patients admitted to the intensive care unit (ICU). The APACHE III prognostic system was invented in 1991 and has been validated in several patient populations. The APACHE III is calculated from clinical data obtained during the first 24 h of ICU admission. It consists of several components, including the primary reason for ICU admission, age, sex, race, preexisting comorbidities, and location prior to ICU admission. Although several modifications have been made to the original APACHE score, only APACHE III has confirmed accuracy for mortality prediction in critically ill HD patients. Dara et al compared the performance of APACHE III with the SOFA score. They found that APACHE III outperformed SOFA score for mortality prediction in critically ill HD patients. In our systematic review, this was the only study that focused on ICU patients.
The prognostic value of serological biomarkers, such as cardiac troponin T (cTnT) and NT-proBNP, for mortality in dialysis patients has been shown in observational studies[21,22]. As noted earlier, these biomarkers highlight the association between cardiac remodeling and death in dialysis patients. cTnT has been shown to predict mortality in PD patients independent of inflammation, residual renal function, and cardiac hypertrophy. This was supported by the recent Netherland Cooperative Study on the Adequacy of Dialysis report, where the predictive power of cTnT for mortality is superior to other risk factors in a mixed population of HD and PD patients. We suggest the use of cTnT for its prognostic value in at least PD patients due to its ease of testing.
Pre-dialysis neutrophil-lymphocyte ratio (NLR) is a novel and strong short-term predictor for all-cause mortality in HD patients. NLR has been shown to have predictive utility in numerous disease states, including acute myocardial infarction and autoimmune disease[74,75]. NLR represents systemic inflammation, nutritional status, and atherosclerosis, which are all prevalent in CKD patients[76,77]. One single-center Japanese cohort revealed that NLR is superior to other generic biomarkers for 1-year survival among patients with ESKD due to diabetic nephropathy. Although the usefulness of NLR has been widely studied in CKD patients, more data on its accuracy in diverse CKD patients are needed.
Skin autofluorescence (SAF) has been studied in healthy and uremic individuals[78-80]. SAF is predictive of microvascular disease progression, cardiovascular events, and nonspecific adverse clinical outcomes in diabetic and CKD patients regardless of dialysis status[79,81-84]. SAF is measured by illuminating approximately 1 cm2 of skin surface with a 300-420 nm light source. SAF is calculated as the ratio between the emission light and reflected excitation light, multiplied by 100. It is speculated that SAF represents the accumulation of advanced glycation end-products that cause collagen and elastin cross-linking, which can translate into increased arterial stiffness[85-87]. In 2019, Mukai et al found that SAF in CKD patients is predictive of adverse clinical outcomes, including death. This study supports the association between SAF, arterial stiffness, and increased mortality in CKD patients.
PP is the difference between systolic and diastolic blood pressure. Rroji et al demonstrated that PP has a significant impact on dialysis patient survival. In their study, patients with pulmonary hypertension had significantly higher PP compared with patients without pulmonary hypertension. The underlying mechanism between increased PP and death in CKD is not fully understood. However, we postulate that increased PP, either from a reduction of diastolic blood pressure or increase of systolic pressure, could increase morbidity and mortality due to reduced coronary perfusion pressure, elevated systemic vascular resistance, or decreased vascular compliance. However, the predictive value of PP in CKD patients needs to be demonstrated in a larger study.
Detrended fluctuation analysis (DFA) of heart rate dynamics is another prognostic marker for PD patients. In recent years, heart rate variability derived from beat-to-beat heart rate dynamic monitors has been used as a surrogate marker of autonomic modulation in an effort to predict patient outcomes[89-91]. DFA is a scaling analysis method to represent the correlation properties of a signal. This form of analysis permits the detection of long-range correlation embedded in non-stationary time series. To date, a few studies have demonstrated that DFA provides significant information on risk of cardiovascular events in heart failure and acute coronary syndrome patients[94,95]. In 2016, Chiang and colleagues showed that lower short-term DFA was predictive of total mortality in PD patients (median follow-up duration of 34 mo). They were the first study to measure and associate autonomic dysregulation with clinical outcomes in ESRD patients. Subsequent studies have shown consistent data obtained from the general population, where reduced DFA was associated with increased mortality via sudden cardiac death. DFA shows potential promise in CKD patients since ESRD is associated with an overactive sympathetic nervous system. This dysregulation results in an amplification of intracellular cyclic AMP (cAMP), which leads to an increase of the action potentials in the sinoatrial node. These pathophysiological changes can be detected in the altered beat-to-beat variability. However, DFA is a relatively new method of clinical analysis, and its application in other patient populations such as HD and non-dialysis CKD remains to be elucidated.