Predictive value of echocardiography and its relation to Kt/V and anthropometric parameters in hemodialysis patients

Abstract

Background: In order to evaluate the predictive value of echocardiograph parameters for mortality of hemodialysis patients and their relation to Kt/V and anthropometry, a prospective, single center study was analyzed post-hoc. Methods: This analysis encompassed 106 patients on maintenance hemodialysis monitored for 108 months from 1996 to 2004. spKt/V was calculated using the Daugirdas formula. Anthropometric measurements included mid-arm muscle measurements (MAMC) and percentage of body fat (%fat). Echocardiography included the estimations of left ventricular wall thickness, dimensions and volumes (EDV, ESV), systolic LV function (ejection fraction - EFLV, fractional shortening - VCF, stroke volume - SV) and diastolic LV function (E/A, VTI-A wave of transmitral flow velocity), left atrial diameter, as well as assessment of clinical and biochemical parameters. The Cox proportional hazard model was used to estimate predictive values of echocardiograph parameters. Results: Kt/V correlated significantly with left ventricular systolic and diastolic volumes and function, septal and posterior wall thickness and left atrium dimension. MAMC and %fat also correlated with many echocardiograph parameters. Multivariate Cox regression selected age [HR 1.07; CI (1.03–1.12); p < 0.01], albumin [HR 0.88; CI (0.79–0.97); p < 0.05] and left atrium dimension – binary [values > 4 cm were marked as “1” and others “0” – HR 3.76; CI (1.56–9.03); p < 0.01] as independent predictors of death. Conclusion: Left atrium dimension was the most important predictor of mortality among the echocardiograph parameters. Many of these parameters were related to Kt/V and anthropometric measurements and could be the combined consequence of hypervolemia and hypertension.

• Adequacy
• anthropometry
• hemodialysis
• Kt/V
• nutrition
• retrospective study
• survival analysis

Introduction

Cardiovascular disease is a common comorbidity in hemodialysis patients and a major cause of mortality.1 Many traditional and non-traditional risk factors have been identified.2 Malnutrition is common in these patients and is related to inflammation and atherosclerosis.3 Dose of dialysis is also related to cardiovascular morbidity and mortality in hemodialysis patients.4 Early detection of cardiovascular comorbidity is important. Echocardiography is a non-invasive cardiac imaging technique that is widely available and enables early detection of abnormalities of cardiac morphology and function.5 However, there is a lack of information about the relation of echocardiography with dose of dialysis and nutrition. In order to elucidate these relations, we investigated the predictive value for mortality of echocardiograph parameters together with their relations to Kt/V and anthropometric factors.

Methods

This post-hoc analysis of an observational study included a cohort of 106 patients (58 men) on maintenance hemodialysis in the Dialysis Unit of the Clinical Center of Serbia.6 The design of the study was approved by the Ethics Committee of the Clinical Center.

The investigation included patients who had undergone echocardiography in the period from 1 January 1996 to 31 December 2002 as the only inclusion criterion and they were monitored until 31 December 2004. Previously they had been on hemodialysis for up to 225 months. Baseline measurements were made during the first year after they entered the study and involved clinical, echocardiograph, anthropometrical (February) and laboratory examinations. Many patients in this cohort did not undergo repeated echocardiograph examination, so only one baseline measurement for each patient was analyzed.

All patients were monitored until their death (42 patients), departure from the center (7 patients), kidney transplantation (2 patients), drop out of the study (2 patients) or until its end on 31 December 2004 (30 men and 23 women).

Echocardiograph measurements included the estimation of left ventricular wall thickness, dimensions and volume (EDV, ESV), systolic LV function (ejection fraction-EFLV, fractional shortening-VCF, stroke volume – SV) and diastolic LV function (E/A, VTI-A wave of transmitral flow velocity), left atrial diameter, as well as assessing clinical and biochemical parameters. Left ventricular mass index (LWMI) was calculated according to recommendations of the American Society of Echocardiography.7 The VTI-A wave of transmitral flow velocity was measured in only 86 patients and peak velocity of pulmonic flow was determined in only 82 patients. Echocardiography was performed on the day between two hemodialysis sessions according to recommendations of the American Society of Echocardiography.8 Normal values included body size, age and gender corrections. All echocardiograph examinations were performed by one cardiologist who was experienced in echocardiograph imaging.

Mean arterial pressure (MAP) was calculated as diastolic pressure – (systolic pressure – diastolic pressure)/3 and the normal interval was 70–100 mmHg. All patients who had edema, moist inspiratory crepitant rales over the lung bases before hemodialysis, signs of hypervolemia on radiologic examination or echotomographic signs of dilated heart chambers and congestive heart failure were recorded as in a hypervolemic state. Patients with clinical, ECG or radiological evidence of ischemic heart disease were also detected.

Fat percentage (%fat) was determined from the sum of triceps (TSF), biceps, suprailiac and subscapular skinfolds, as recommended by the DOQI Clinical Practice Guidelines.9 Normal intervals were 12–20% for men and 20–30% for women.10 The BMI normal interval was 18.6–24.9 kg/m².11 Mid-arm muscle circumference (MAMC) was calculated from mid-arm circumference (MAC) using the formula [MAMC (cm) = MAC (cm) − 0.314 * TSF (mm)].

The patients were dialyzed thrice per week for 3–5 h using bicarbonate dialysis. They were prescribed a standard diet for hemodialysis patients12 that did not change during the observation period. The dialyzers were never reused. Kt/V was calculated using the second generation Daugirdas formula.13 where R = post-dialysis/pre-dialysis blood urea nitrogen, t = dialysis hours, UF = pre–post dialysis weight change and W = post-dialysis weight.

Statistical methods

The outcome variable in Cox’s regression was survival time in months. Predictor variables used in Cox’s regression were derived from data: (1) in the baseline period and (2) at entry into the study.

Variables obtained in the baseline period: Echocardiograph and anthropometrical parameters were measured only once during the baseline period (February). Individual mean values of laboratory parameters, Kt/V from monthly measurements in the 12-month baseline period and a separate variable were constructed for each parameter. Comorbidity was determined during the baseline period and single binary predictor variables were used for ischemic heart disease and hypervolemic state.

Variables obtained at entry into the study: The following predictor variables were determined when the patients entered the study: age (years), gender (male/female) and time of dialysis (months). All binary variables were coded as 0/1 except gender (coded 1/2). The Kolmogorov–Smirnov test was used for testing distributions of the variables.

Pearson’s and Spearman’s correlation coefficients were employed for detecting associations among predictor variables. Some variables were collinear (correlation coefficient > 0.90) such as the echocardiograph parameters: diastolic dimension and end diastolic volume of the left ventricle, systolic dimension and end systolic volume of left ventricle, ejection fraction and fractional shortening. Many echocardiograph parameters were possibly collinear (correlation coefficient >0.50), such as aortic peak velocity and peak velocity of pulmonic flow, systolic and diastolic dimensions of the left ventricle, posterior and septal wall thickness. In addition, left ventricular mass index was possibly collinear with posterior wall thickness, septal wall thickness, diastolic dimension of the left ventricle and cardiac index. Other possibly collinear variables were: systolic dimension and end diastolic volume of the left ventricle, diastolic dimension and end systolic volume of the left ventricle, end systolic and end diastolic dimension of the left ventricle, systolic dimension of the left ventricle and stroke volume, end diastolic volume of the left ventricle and stroke volume, end systolic volume of the left ventricle and stroke volume, cardiac output and posterior wall thickness, cardiac output and stroke volume, ejection fraction and systolic dimension of the left ventricle, ejection fraction and end systolic volume of the left ventricle, fractional shortening and systolic dimension of the left ventricle, fractional shortening and end diastolic volume of the left ventricle, cardiac index and diastolic dimension of the left ventricle, cardiac index and end diastolic volume of the left ventricle, cardiac index and end systolic volume of the left ventricle, cardiac index and stroke volume, stroke index and diastolic dimension of the left ventricle, stroke index and end diastolic volume of the left ventricle, stroke index and stroke volume, stroke index and cardiac index, mean velocity of circumferential fiber shortening and end diastolic volume of the left ventricle, mean velocity of circumferential fiber shortening and ejection fraction, mean velocity of circumferential fiber shortening and fractional shortening, flow duration and deceleration time, VTI A wave of mitral flow and mean velocity of circumferential fiber shortening, velocity time integral of mitral flow and VTI A wave of mitral flow, acceleration time and end systolic volume of the left ventricle. Correlation among echocardiograph parameters with coefficients <0.5 (that were not possibly collinear) are not presented. Correlations among clinical parameters were not collinear (coefficients were <0.5 and are not presented).

Univariate survival analysis was performed with the Cox proportional hazard model. The primary dependent variable was the time to death measured in months. Variables that were potential predictors of death in univariate analysis (p < 0.10) were tested in the multivariate Cox proportional hazard model using the stepwise method. Proportional hazards were tested by fitting time-varying models.

Results

Demographic data and the primary kidney disease for all patients at the time of inclusion in the study are presented in Table 1. Glomerulonephritis was the most frequent disease, while hypertension and diabetes were together responsible for end-stage renal disease in 15.1% of the patients.

Table 1. Demographic variables, diagnosis, clinical and laboratory data.

Parameter	Value
Age (years)^a	53.3 (12.2)
Dialysis vintage (months)^b	53.0 (24.7–97.7)
Sex (men/women)	58/48
Primary renal diagnosis
Nephrosclerosis (hypertension)	10.4%
Diabetes	4.7%
Glomerulonephritis	17.9%
Polycystic kidney disease	13.2%
Interstitial nephritis	8.5%
Other	22.6%
Unknown	22.6%
Kt/V^a	1.21 (0.22)
MAP (mmHg)^a	97 (9)
Ischemic heart disease	26.4%
Hypervolemic state	40.5%
Laboratory analysis
Hemoglobin (g/l)^a	93.3 (16.0)
Urea (mmol/l)^a	28.0 (4.1)
Creatinine (µmol/l)^a	1047 (203)
Calcium (mmol/l)^a	2.29 (0.20)
Phosphorus (mmol/l)^a	1.85 (0.36)
Albumin (g/l)^a	39.9 (4.7)

Notes: ^aMean (SD); ^bmedian (interquaritle value).

Laboratory and clinical data are shown in Table 1. Cardiovascular comorbid conditions were very frequent as can be seen for ischemic heart disease and hypervolemic state. In addition, 50 patients (47%) had MAP >100 mmHg and only three had MAP >110 mmHg. Patients had Kt/V and hemoglobin values lower than currently recommended.

Echocardiograph parameters are shown in Table 2. The mean values of left ventricular mass index were higher than upper normal value as well as stroke volume and isovolumetric relaxation time. Mean values of E/A, median of acceleration and deceleration time were lower than normal values. VTI A wave of mitral flow/VTI of mitral flow were increased. The percentages of deviation from normal values were high for all echocardiograph parameters (Table 2). For example, LA diameter greater than 4 cm was seen in 26% and E/A lower than 1 in 91% of the patients.

Table 2. Echocardiograph parameters.

Parameter	Mean (SD) or median (interquartile value)	Percentage of deviation from normal values
Peek velocity of tricuspid flow (m/s)^a	0.63 (0.10)	13%
Aortic dimension (cm)^a	3.35 (0.45)	17%
Systolic separation of aortic leaflet (cm)^a	1.88 (0.33)	11%
Aortic peek velocity (m/s)^a	1.50 (0.65)	47%
Peak velocity of pulmonic flow (m/s)^a	1.09 (0.23)	45%
LV diastolic dimension (cm)^a	5.18 (0.75)	22%
LV systolic dimension (cm)^a	3.47 (0.70)	17%
Septal thickness (cm)^a	1.04 (0.22)	25%
Posterior wall thickness (cm)^a	0.95 (0.15)	10%
Left ventricular mass index (g/m^2 bsa)^a	116 (44)	64%
LA dimension (cm)^a	3.72 (0.61)	26%
RV diastolic dimension (cm)^b	2.31 (0.55)	16%
LV end diastolic volume (ml)^b	125 (101–145)	32%
LV end systolic volume (ml)^b	49.0 (34.5–61.0)	38%
Stroke volume (ml)^a	78.2 (27.3)	53%
Cardiac output (l/min)^a	5.66 (1.82)	19%
Ejection fraction (%)^a	60.3 (9.7)	25%
Fractional shortening (%)^a	31.8 (6.7)	33%
Cardiac index (l/min/m^2BSA)^a	3.49 (1.29)	46%
Stroke index (ml/m^2BSA)^a	43.7 (17.9)	55%
Mean velocity of circumf. fiber shortening (Vcf, circ/s)^a	1.33 (0.43)	15%
Acceleration of transmitral flow^b	11.2 (7.5–18.2)	56%
Acceleration time (m/s^2)^b	50.0 (40.0–70.0)	84%
Deceleration of transmitral flow (m/s^2)^b	5.19 (2.23–16.07)	63%
Deceleration time (ms)^a	119 (47)	76%
Flow duration (ms)^a	325 (105)	14%
Velocity time integral of mitral flow (m)^a	0.14 (0.05)	40%
VTI A wave of mitral flow (m)^b	0.06 (0.05–0.09)	18%
Isovolumetric relaxation time (ms)^a	126 (29)	88%
Duration of A wave of mitral flow (ms)^a	126 (39)	28%
E/A^a	0.69 (0.20)	91%

Notes: LV – left ventricular; RV – right ventricular; LA – left atrium; BSA – body surface area.

^aMean (SD), normal distribution (Kolmogorov–Smirnov test); ^bmedian (interquartile value).

Anthropometric parameters are shown in Table 3 separately for men and women. Although SDs for triceps skinfold and percentage of body fat were large, the Kolmogorov–Smirnov test did not find significant differences from normal distribution.

Table 3. Anthropometric variables.

Parameter	Men	Women
Body weight (kg)	72.0 (14.0)	58.6 (11.1)
BMI (kg/cm2)	23.6 (4.0)	22.7 (3.7)
Mid-arm circumference (cm)	28.6 (3.9)	26.8 (4.8)
Mid-arm muscle circumference (cm)	24.9 (2.6)	22.3 (3.3)
Triceps skinfold (mm)	11.6 (6.6)	14.2 (6.0)
Percentage of body fat	20.9 (7.8)	28.9 (6.9)

Note: All variables were expressed as mean (SD).

Men had significantly lower Kt/V when compared to women (t-test, p < 0.01). In addition, the t-test showed that men had significantly greater aortic dimension, systolic separation of the aortic leaflet, systolic and diastolic dimensions of the left ventricle (p < 0.05) but a lower ejection fraction (p < 0.01). End systolic and diastolic volumes of the left ventricle were also significantly greater in men (Mann–Whitney test, p < 0.01).

Patients with normal left ventricular mass index (M < 115, F < 95) had significantly lower MAP and higher hemoglobin levels (t-test, p < 0.05). Patients with normal left atrial dimension (<4 cm) had significantly higher albumin concentrations and Kt/V but lower phosphate levels (t-test, p < 0.05). There was no significant difference in dialysis duration between patients with normal and large left atrium dimensions.

Patients with Kt/V > 1.2 had significantly lower aortic dimensions, systolic separation of the aortic leaflet, peak velocity of pulmonic flow, LV diastolic dimension, LV systolic dimension, septal thickness, left atrial dimension, LV end diastolic volume, LV end systolic volume, stroke volume, and significantly higher ejection fraction and fractional shortening (Table 4). These patients had lower cardiac output, higher RV diastolic dimension, higher mean velocity of circumferential fiber shortening and lower posterior wall thickness but the differences only approached statistical significance (p < 0.10). In addition, patients with Kt/V > 1.2 had significantly higher albumin (t-test, p < 0.05) and significantly lower MAP (t-test, p < 0.05), phosphate (t-test, p < 0.05), creatinine (t-test, p < 0.05) and MAMC (t-test, p < 0.01). There were no significant differences in hemoglobin levels and percentage of body fat. The group of patients with a hypervolemic state had slightly lower levels of Kt/V but the difference was not significant.

Table 4. Differences between Kt/V subgroups and echocardiograph parameters and between fat percentage subgroups and echocardiograph parameters.

Parameter	Kt/V ≤ 1.2	Kt/V > 1.2	%fat ≤50 pct	%fat >50 pct
Peek velocity of tricuspid flow (m/s)^a	0.62	0.63	0.62	0.63
Aortic dimension (mm)^a	3.48**	3.21**	3.44*	3.26*
Systolic separation of aortic leaflet (cm)^a	1.97**	1.79**	1.83	1.93
Aortic peek velocity (m/s)^a	1.56	1.44	1.61	1.40
Peak velocity of pulmonic flow (m/s)^a	1.14*	1.03*	1.07	1.10
LV diastolic dimension (cm)^a	5.47**	4.83**	5.07	5.28
LV systolic dimension (cm)^a	3.76**	3.13**	3.40	3.54
Septal thickness (cm)^a	1.10**	0.96**	1.00	1.08
Posterior wall thickness (cm)^a	0.98	0.92	0.95	0.95
Left ventricular mass index (g/m^2 bsa)^a	123	109	121	111
LA dimension (cm)^a	3.83*	3.58*	3.72	3.72
RV diastolic dimension (cm)^b	2.22	2.43	2.32	2.30
LV end diastolic volume (ml)^b	149**	112**	127	135
LV end systolic volume (ml)^b	63.2**	40.9**	50.9	54.2
Stroke volume (ml)^a	85.2**	70.2**	76.5	79.8
Cardiac output (l/min)^a	6.03	5.29	5.18*	6.11*
Ejection fraction (%)^a	57.5**	63.5**	60.1	60.6
Fractional shortening (%)^a	30.4*	33.5*	31.9	31.8
Cardiac index (l/min/m^2BSA)^a	3.59	3.37	3.58	3.42
Stroke index (ml/m^2BSA)^a	45.2	42.1	41.5	45.7
Mean velocity of circumf. fiber shortening (Vcf, circ/s)^a	1.21*	1.48*	1.25	1.40
Acceleration of transmitral flow^b	14.4	16.9	15.0	16.0
Acceleration time (m/s^2)b	65.7	57.5	58.7	65.2
Deceleration of transmitral flow (m/s^2)b	38.4	27.1	29.8	36.7
Deceleration time (ms)^a	116	123	126	113
Flow duration (ms)^a	323	329	328	323
Velocity time integral of mitral flow (m)^a	0.14	0.14	0.15	0.13
VTI A wave of mitral flow (m)^b	0.10	0.07	0.06	0.11
Isovolumetric relaxation time (ms)^a	133	119	123	130
Duration of A wave of mitral flow (ms)^a	126	126	127	126
E/A^a	0.71	0.67	0.72	0.66

Notes: LV – left ventricular; RV – right ventricular; LA – left atrium; BSA – body surface area.

^at- test; ^bMann–Whitney test.

*p < 0.05; **p < 0.01; 50 pct – 50 percentile value separately for men and women.

According to percentage of body fat, overweight patients exhibited a significantly higher cardiac output (t-test, p < 0.05), while underweight patients had faster mean velocity of circumferential fiber shortening (t-test, p < 0.01). Differences in the echocardiograph parameters between groups made according to the 50 percentile value of percentage of body fat separately determined for men and women are displayed in Table 4. Groups made according to the 50 percentile value of mid-arm muscle circumference revealed only nearly significant differences for several echocardiograph parameters.

Significant correlations between Kt/V and many echocardiograph parameters were found. Thus, Kt/V correlate with aortic dimension (Pearson, r = −0.29; p < 0.01), LV diastolic dimension (Pearson, r = −0.46; p < 0.01), LV systolic dimension (Pearson, r = −0.47; p < 0.01), septal thickness (Pearson, r = −0.29; p < 0.05), posterior wall thickness (Pearson, r = −0.28; p < 0.05), left atrial dimension (Pearson, r = −0.24; p < 0.05), LV end diastolic volume (Spearman, ρ = −0.47; p < 0.01), LV end systolic volume (Spearman, ρ = −0.46; p < 0.01), stroke volume (Pearson, r = −0.32; p < 0.01), ejection fraction (Pearson, r = 0.30; p < 0.01) and fractional shortening (Pearson, r = 0.24; p < 0.05).

Significant correlations between MAP and many echocardiograph parameters were also found. Thus, MAP correlated with LV diastolic dimension (Pearson, r = 0.37; p < 0.01), LV systolic dimension (Pearson, r = 0.27; p < 0.01), septal thickness (Pearson, r = 0.33; p < 0.05), posterior wall thickness (Pearson, r = 0.35; p < 0.01), LV mass index (Pearson, r = 0.43; p < 0.01), left atrial dimension (Pearson, r = 0.24; p < 0.05), LV end diastolic volume (Spearman, ρ = 0.30; p < 0.01), LV end systolic volume (Spearman, ρ = 0.27; p < 0.01) and stroke volume (Pearson, r = 0.33; p < 0.01). The correlation between Kt/V and MAP was weak (Pearson, r = −0.18; p < 0.06).

Besides Kt/V and MAP, significant correlations between many echocardiograph and other clinical parameters (dialysis duration, age, creatinine, etc.) were also found. Thus, LV diastolic dimension (Pearson, r = −0.30; p < 0.01), LV systolic dimension (Pearson, r = −0.25; p < 0.01) and stroke index (Pearson, r = −0.37; p < 0.01) correlated with dialysis duration. Aortic dimension (Pearson, r = −0.32; p < 0.01), peak velocity of pulmonic flow (Pearson, r = −0.47; p < 0.01), septal thickness (Pearson, r = −0.27; p < 0.05), RV diastolic dimension (Pearson, r = 0.31; p < 0.01) and fractional shortening (Pearson, r = −0.22; p < 0.05) correlated significantly with age. LV systolic dimension showed a significant correlation with albumin (Pearson, r = −0.21; p < 0.05). Aortic dimension, systolic separation of the aortic leaflet, end diastolic and systolic volumes of LV and stroke volume were positively correlated with creatinine, while VTI A wave of mitral flow were positively correlated with serum calcium. Only flow duration significantly negatively correlated with phosphate. No one echocardiograph parameter correlated significantly with hemoglobin or urea.

Correlation analysis revealed significant associations between many echocardiograph and anthropometric parameters. Body weight correlated with aortic dimension (Pearson, r = 0.23; p < 0.05), systolic separation of aortic leaflet (Pearson, r = 0.35; p < 0.01), LV diastolic dimension (Pearson, r = 0.34; p < 0.01), LV systolic dimension (Pearson, r = 0.32; p < 0.01), end diastolic volume LV (Spearman, ρ = 0.41; p < 0.01), end systolic volume LV (Spearman, ρ = 0.34; p < 0.01), stroke volume (Pearson, r = 0.28; p < 0.01) and cardiac output (Pearson, r = 0.26; p < 0.05).

BMI correlated with systolic separation of aortic leaflet (Pearson, r = 0.23; p < 0.05), LV diastolic dimension (Pearson, r = 0.30; p < 0.01), LV systolic dimension (Pearson, r = 0.23; p < 0.05), end diastolic volume LV (Spearman, ρ = 0.23; p < 0.05), stroke volume (Pearson, r = 0.26; p < 0.05) and cardiac output (Pearson, r = 0.23; p < 0.05) and VTI A wave of mitral flow (ρ = 0.38; p < 0.05).

Mid-arm circumference correlated with systolic separation of aortic leaflet (Pearson, r = 0.27; p < 0.01), LV diastolic dimension (Pearson, r = 0.28; p < 0.01), LV systolic dimension (Pearson, r = 0.24; p < 0.05), end diastolic volume LV (Spearman, ρ = 0.31; p < 0.01), end systolic volume LV (Spearman, ρ = 0.24; p < 0.05), stroke volume (Pearson, r = 0.23; p < 0.05), cardiac output (Pearson, r = 0.29; p < 0.05), and VTI A wave of mitral flow (ρ = 0.49; p < 0.01).

Mid-arm muscle circumference correlated with systolic separation of aortic leaflet (Pearson, r = 0.30; p < 0.01), LV diastolic dimension (Pearson, r = 0.34; p < 0.01), LV systolic dimension (Pearson, r = 0.33; p < 0.01), end diastolic volume LV (Spearman, ρ = 0.45; p < 0.01), end systolic volume LV (Spearman, ρ = 0.40; p < 0.01), stroke volume (Pearson, r = 0.28; p < 0.01) and cardiac output (Pearson, r = 0.29; p < 0.05).

Triceps skinfold correlated with aortic dimension (Pearson, r = −0.27; p < 0.01), aortic peek velocity (Pearson, r = −0.25; p < 0.05) and VTI A wave of mitral flow (Spearman, ρ = 0.46; p < 0.01).

Percentage of body fat correlated with aortic dimension (Pearson, r = −0.35; p < 0.01), and mean velocity of circumferential fiber shortening (Pearson, r = 0.34; p < 0.05).

Mortality rate was 39.6% (42/106) during the observation period. Potential predictors of death were selected by the univariate Cox proportional hazard model (p < 0.10, Table 5). No anthropometric parameter was selected as a potential predictor of death. Because left atrium diameter was non-proportional, a new binary variable was formed where values greater than 4 cm were marked as “1” and the others as “0”. Similarly, values of Kt/V less than 1.2 were marked as “0” and the others “1” because Kt/V was non-proportional. However, the new variable was also not proportional, so a Kt/V binary variable with a cut-off at 1.3 was used. Among echocardiograph parameters, aortic peak velocity, systolic dimension of the left ventricle, left atrium dimension – binary and ejection fraction were selected for further testing. End diastolic and systolic volumes of the left ventricle were excluded because of collinearity with ejection fraction, as well as diastolic dimension because of collinearity with the systolic dimension of the left ventricle. Unfortunately, peak velocity of pulmonic flow and VTI A wave of mitral flow also had to be excluded due to the large number of missing values.

Table 5. Cox proportional hazard model – selected parameters p < 0.10.

Parameter	Hazard ratio (CI)	p
Age	1.04 (1.01–1.07)	<0.01
Nephrosclerosis	2.31 (1.06–5.00)	0.03
Glomerulonephritis	0.43 (0.16–1.10)	0.08
Ischemic heart disease	1.82 (0.96–3.43)	0.06
Kt/V*	0.08 (0.01–0.39)	<0.01
Kt/V – binary (0 values <1.2 other 1)^a	0.37 (0.19–0.73)	<0.01
Kt/V – binary (0 values <1.3 other 1)	0.36 (0.16–0.79)	0.01
Urea	1.08 (1.00–1.17)	0.04
Albumin	0.87 (0.80–0.93)	<0.01
Aortic peek velocity	2.88 (1.70–4.78)	<0.01
Peak velocity of pulmonic flow^b	10.54 (1.53–72.29)	0.01
LV diastolic dimension	1.55 (1.03–2.34)	0.03
LV systolic dimension^c	1.67 (1.06–2.61)	0.02
LV mass index^a	1.00 (1.00–1.01)	0.02
LA dimension^a	1.78 (1.12–2.83)	0.01
LA dimension - binary (0 values < 4 cm other 1)	2.94 (1.52–5.69)	<0.01
End diastolic volume LV^d	1.00 (1.00–1.01)	0.01
End systolic volume LV^d	1.01 (1.00–1.02)	<0.01
Ejection fraction	0.97 (0.94–1.00)	0.07
VTI A wave of mitral flow^b	803.42 (5.95–108397.70)	<0.01

Notes: Univariate test of parameters, all-cause mortality in the cohort.

^aExcluded – no proportional; ^bExcluded – missing values; ^cExcluded – collinear with ejection fraction; ^dExcluded – collinear with LV systolic dimension; Hazard ratio – Exp (B), CI – 95% confidence interval of Exp (B), p – significance of coefficient.

To confirm the independent predictive power for patient mortality, the above mentioned data from Table 5 were entered into the multivariate Cox proportional hazard model (Table 6). Age, albumin and left atrium dimension – binary were selected as independent predictors of all-cause mortality. The risk for patient mortality was 3.76 times greater for patients who had an enlarged left atrium when it was adjusted for other risk factors.

Table 6. Predictors of all-cause mortality in the cohort.

Parameter	Hazard ratio (CI)	p
Age	1.07 (1.03–1.12)	<0.01
Albumin	0.88 (0.79–0.97)	0.01
LA dimension – binary  (0 values <4 cm other 1)	3.76 (1.56–9.03)	<0.01

Note: Hazard ratio – Exp (B), CI – 95% confidence interval of Exp (B), p – significance of coefficient. Multivariate Cox proportional hazard model.

Discussion

The present study found that patients with Kt/V > 1.2 had significantly lower aortic dimension, systolic separation of the aortic leaflet, peak velocity of pulmonic flow, LV diastolic and systolic dimensions, septal thickness, left atrial dimension, LV end diastolic and systolic volumes, stroke volume, together with significantly higher ejection fraction and fractional shortening. Correlation analysis revealed significant correlations between Kt/V and the previous variables. In addition, patients with Kt/V > 1.2 had significantly lower MAP but the correlation between Kt/V and MAP was weak. On the other hand, anthropometric parameters in group analysis revealed few significant differences but correlation analysis found many significant associations. Survival analysis revealed age, nephrosclerosis, glomerulonephritis, ischemic heart disease, Kt/V, urea, albumin, aortic peak velocity, peak velocity of pulmonic flow, LV diastolic dimension, LV systolic dimension, LA dimension, LV end diastolic volume, LV end systolic volume, ejection fraction and VTI A wave of mitral flow as potential predictors of mortality. Multivariate Cox regression selected age, albumin and LA dimension (binary) as independent predictors of mortality.

Hypertension and hypervolemia are very frequent in hemodialysis patients and therefore many echocardiograph parameters could be abnormal. In addition, the arterio-venous fistula represents a left–right shunt. According to their MAP, many of our patients had high blood pressure and hypertension accompanied primary kidney disease in 10.4% of patients. In addition, 46% of patients expressed hypervolemic state. Abnormal values of echocardiograph parameters were found in high percentage of our patients but only a few mean values for echocardiograph parameters were abnormal. Peak velocity of tricuspid flow was higher than normal due to the left to right shunt and hypervolemia. VTI A wave of mitral flow/VTI of mitral flow was higher because of hypervolemia and increased left ventricular filling pressure. Low values of E/A represented LV diastolic dysfunction. Among parameters where mean values were within the normal interval, patients with abnormal values for echocardiograph parameters had a worse prognosis.

Echocardiograph parameters in hemodialysis patients are well studied.14 Thus, Avenatti et al.15 determined echocardiograph parameters before, after 90 min and at the end of the dialysis. Palecek et al.16 investigated the effect of preload reduction in hemodialysis patients on echocardiograph parameters. Echocardiograph methods have frequently been used to estimate specific cardiac structure or function. For example, left ventricular mass index and ejection fraction were employed to assess the impact of low triiodothyronine levels on mortality in hemodialysis patients.17 Similarly, echocardiograph measurements were used to estimate cardiac output in an investigation exploring regional blood flow.18 There are many studies on the predictive value of echocardiograph parameters for mortality of hemodialysis patients.19^,20 Kainz et al.21 found that left atrial diameter was a predictor of mortality among renal allograft recipients. Patients with widely enlarged left atria exhibited a considerably reduced life expectancy. Our results are in accordance with these studies. We also found that a large left atrial diameter is an independent predictor of mortality of hemodialysis patients. Thereby, we not only supported earlier findings but also confirmed the reality and validity of our study.

Many echocardiograph parameters were related to dose of dialysis which has rarely been studied before.22 Cao et al. reported that only left ventricular hypertrophy is related to Kt/V in Chinese maintenance hemodialysis patients. We also found higher LV mass index than normal in our patients but there were no significant correlations between Kt/V and LV mass index. MAP was associated with LV mass index in our patients and 47% of them had MAP > 100 mm Hg and many of them (46%) had hypervolemia. Many other parameters were related to Kt/V in our study probably as a consequence of hypervolemia. For example, the relation between Kt/V and systolic separation of the aortic leaflet could be the consequence of high blood pressure and hypervolemia in the lower Kt/V group. In addition, the relation between Kt/V and MAP in our patients was weak and non-significant, so hypervolemia might be a more important parameter related to Kt/V than hypertension. Another example is the relation of Kt/V and aortic dimension which could also be caused by hypervolemia. However, the difference in hypervolemia between the lower and normal Kt/V groups was also non-significant, so evidence of the relation between Kt/V and hypervolemia was indirect as for hypertension. One could expect a greater effect of hypervolemia on cardiac function in our patients because 46% exhibited hypervolemia but this is mainly intermittent in nature because patients wax between dialysis and wane during dialysis.

A number of echocardiograph parameters correlated positively with anthropometric parameters. This could be explained simply by larger body mass and the consequently enlarged capillary bed and blood volume. Besides nonspecific anthropometric parameters such as body weight and body mass index, there are specific anthropometric parameters that determine different body compartments. For example, percentage of body fat and triceps skinfold measure body fat content and mid arm muscle circumference measures muscle mass of the body. Therefore, there were differences in relations between anthropometric and echocardiograph parameters. Some relations could not be explained simply, such as the correlation between VTI A wave of mitral flow and triceps skinfold but no correlation of VTI wave of mitral flow and percentage of body fat. When we investigated groups of overweight and underweight patients and their relation to echocardiograph parameters, only a few associations were significant. Although, correlation analysis only revealed weak coefficients between echocardiograph and anthropometric parameters, this is an important finding because connections between cardiovascular morbidity and nutrition gained importance after the Framingham study. Relations between cardiovascular comorbidity and nutrition using echocardiography in hemodialysis patients have rarely been examined but without anthropometry.22^,23 However, anthropometric measurements were more specific than body mass index or subjective global assessment.

Correlations between many echocardiographs and a number of clinical parameters (dialysis duration, age, creatinine, etc.) could also be explained by hypervolemia and the effect of calcium on the myocardium.

This study has a few drawbacks. First of all, almost half the patients in the baseline period and during the observation period had Kt/V < 1.2 in this post-hoc analysis. Target Kt/V was higher than 1.2 but was not possible to attain because of economic depression.24 However, this enabled an analysis of the relation between echocardiograph parameters and Kt/V. Hemoglobin values were lower than normal due to a lack of erythropoietin stimulating agents but hemoglobin was not a potential predictor of mortality of our patients. Another drawback is the relatively small number of patients for survival analysis, but confidence intervals were narrow and the analysis reliable. In addition, our most important finding is the relation connecting echocardiograph parameters with Kt/V and anthropometric parameters.

In conclusion, left atrium dimension was the most important predictor of mortality among the echocardiograph parameters. Many of them were related to Kt/V and anthropometric measurements. An enlarged left atrium could be the consequence of hypervolemia and a lower dialysis dose. Besides dialysis dose, hypervolemia depends on nutrition and, together with hypertension, could be a cause of worsening left ventricular function.

Acknowledgments

Data presented in this study are obtained from the doctoral dissertation of Dr Milan Stosovic entitled “Analysis of different dialysis adequacy indices based on clinical parameters upon different dialysis doses in hemodialysis patients” approved by the Scientific Committee, School of Medicine, University of Belgrade, IRB vote number 7/5, 21 April 1992.

Declaration of interest

This work was supported by the Ministry of Education and Science of Serbia, contract 175089.