Hemodynamic phenotypes in chronic kidney disease patients based on linear regression of blood pressure parameters

Abstract

Classic and non‐classic cardiovascular (CV) risk factors accumulate in chronic kidney disease (CKD), contributing to vascular remodeling and hemodynamic abnormalities. This study aimed to determine hemodynamic phenotypes based on linear regression of blood pressure (BP) parameters in stage G3‐G4 CKD patients at very high CV risk. 24‐h ambulatory BP monitoring (ABPM), carotid‐femoral pulse wave velocity (PWV) and central BP were obtained from 52 patients (aged 60 ± 11 years, BMI 30 ± 6 kg/m²) with stage G3‐G4 CKD (eGFR 44 ± 12 mL/min./1.73 m²). Linear BP regression coefficients were generated to determine hemodynamic phenotypes using ABPM data. Coexisting hypertension was present in 45 (86%) patients, out of whom 33 (73%) had BP controlled. 24‐h mean systolic/diastolic BP was 128 ± 18/75 ± 12 mm Hg. Twenty‐six patients demonstrated the harmonious (H) and 26 patients diastolic dysfunctional (D) hemodynamic phenotypes. eGFR was not significantly different between both phenotypes. Compared to phenotype H, patients with phenotype D were older (57 ± 11 vs. 63 ± 10 years, p = .04), had higher PWV (8.2 [7.3–10.3] vs. 9.7 [8.3–10.9] m/s, p = .02), ambulatory arterial stiffness index (AASI) (0.31 ± 0.1 vs. 0.40 ± 0.1, p = .02), systolic BP (128 [122–130] vs. 137 [130–150] mm Hg, p = .001) and systolic BP variability (BPV) (11.7 ± 2.3 vs. 15.7 ± 3.4 mm Hg, p < .0001). Our findings suggest that one in two patients with stage G3‐G4 CKD demonstrates an unfavorable D hemodynamic phenotype based on a linear regression model, associated with higher PWV, AASI, systolic BP, and systolic BPV. Further studies are required to assess the clinical utility of hemodynamic phenotypes and whether the D phenotype may predict latent circulatory disorders and outcomes.

1. INTRODUCTION

The global prevalence of chronic kidney disease (CKD) has been estimated at 9.1%, equivalent to 700 million people. ¹ Among CKD patients, the likelihood of a cardiovascular (CV) event increases with disease progression due to the accumulation of classical and non‐classical CV risk factors. ² , ³ Early vascular ageing occurs during CKD, reflected in increased arterial stiffness and abnormal hemodynamic conditions. ⁴ , ⁵ , ⁶ This leads to a vicious circle – on the one hand, impaired renal function promotes vascular sclerosis, and on the other hand, disturbed hemodynamic conditions promote further progression of the disease. As renal function declines, the incidence of vascular calcification increases and occurs many years earlier in CKD patients than in the general population, ⁷ and is associated with several adverse clinical outcomes including ischemic cardiac events, and all‐cause and CV mortality. ² , ⁸ While the progression of CKD involves multiple pathophysiological pathways, the activation of the sympathetic nervous system plays a key role in the development and progression of CKD, ³ , ⁹ arterial stiffening, ¹⁰ , ¹¹ and predicted CV mortality in patients with end‐stage renal disease. ² , ¹² Consequently, patients in all stages of CKD are at higher risk of CV diseases compared to the general population ¹³ and patients with hypertension (HT) and preserved kidney function. ¹⁴ Among CKD patients, the risk of all‐cause and CV mortality increases with declining renal function. ¹⁵ The probability of a CV event is ten times higher in stage G3 and 50 times higher in stage G4 in comparison to those with preserved kidney function. ² Thus, this high CV risk group may benefit from the diagnosis of latent circulatory disorders occurring even in patients with apparently satisfactory blood pressure (BP) control.

A variety of arterial stiffness and hemodynamics measures have been used in clinical and research settings, including pulse wave velocity (PWV), central BP, ambulatory arterial stiffness index (AASI), BP variability (BPV), and pulse pressure (PP) components, among others. ¹⁶ , ¹⁷ , ¹⁸ While PWV is the gold standard in examining arterial stiffness, ¹⁹ the method is not widely used in clinical practice due to the limited availability of equipment and varying results, which depend on heart rate (HR) and BP. ²⁰ Therefore, other methods are being sought to identify patients at increased risk of CV diseases in a noninvasive and reproducible manner.

Previous studies found abnormalities in hemodynamic phenotypes in patients with apparently well‐controlled HT or obstructive sleep apnea (OSA) using a linear regression model of numerical data derived from ambulatory BP monitoring (ABPM). ²¹ , ²² , ²³ This study aimed to assess the hemodynamic phenotypes based on a linear regression model of BP in stage G3‐G4 CKD patients.

2. MATERIALS AND METHODS

2.1. Study participants and protocol

Adult patients were recruited in the Outpatient Nephrology Clinic of the University Clinical Centre of the Medical University of Gdansk, Poland between 2019 and 2021. Inclusion criteria were: diagnosis of stage G3 or G4 CKD according to the KDIGO classification (i.e., estimated glomerular filtration rate (eGFR) between 59 and 45 mL/min/1.73 m² for stage G3a, 44 and 30 mL/min/1.73 m² for stage G3b and between 29 and 15 mL/min/1.73 m² for stage G4, persisting for at least 3 months), ²⁴ stable clinical conditions within 8 weeks preceding the study enrollment (i.e., no exacerbation of chronic conditions nor acute disease such as an infection) with unchanged BP‐lowering medications. Exclusion criteria were: acute kidney injury, past stroke, transient ischemic attack, symptomatic atherosclerotic CV disease (i.e., coronary artery disease, peripheral artery disease), cardiac arrhythmia, atrial fibrillation, autoimmune disease, vasculitis, long‐standing (i.e., more than 5 years) and/or uncontrolled diabetes, OSA, chronic obstructive pulmonary disease, plasma hemoglobin level below 10 g/dL, erythropoietin therapy, kidney transplantation, and nephrectomy.

One hundred and fifteen of 543 patients met the study inclusion criteria. Patients were contacted by phone and 52 were qualified based on their medical history and consent to participate in the study.

All patients underwent a single outpatient clinic visit at the University Translational Medicine Centre, during which medical history, physical examination, anthropometric measurements and routine blood tests were obtained. At the same visit, office BP measurement, electrocardiogram and PWV measurements were performed. At the end of the visit, 24‐h ABPM was performed.

2.2. Laboratory examinations

Fasting blood tests including full blood count, serum creatinine, sodium and potassium levels were performed in the Central Diagnostic Laboratory of the University Clinical Centre hospital using an Abbot Architect analyzer. eGFR based on the creatinine concentration was calculated according to the CKD Epidemiology Collaboration (CKD‐EPI) formula. ²⁵

2.3. Office and 24‐h ABPM

Office BP was measured simultaneously on both arms in a seated position, after 5 min of rest, using a validated Microlife WatchBP device (Microlife AG Swiss Corporation, Switzerland). Three simultaneous measurements on both arms were performed at 1‐min intervals. The final values of systolic BP (SBP) and diastolic BP (DBP) on the right and left arm were calculated as an average of the three measurements performed on each arm. Values from the arm with higher average BP readings were used for analysis.

Twenty‐four‐hour ABPM was performed with the validated Spacelabs ABP 90207 recorder (Spacelabs Healthcare, Redmond, WA, USA). ²⁶ Measurements were obtained every 20 min throughout the day and every 30 min at night. Daytime was defined as the interval between 6:00 a.m. and 09:59 p.m. and nighttime as the interval between 10:00 p.m. and 05:59 a.m. Patients were asked to maintain their regular activity during the day and to adapt the time of night rest to the above intervals. If this condition was not fulfilled, ABPM was repeated. The same was done if less than 70% of the planned ABPM readings were valid. Non‐dipper profile was defined if mean nighttime SBP/DBP was not at least 10% lower than mean daytime SBP/DBP.

2.4. Pulse wave velocity and pulse wave analysis

PWV measurement was performed using the SphygmoCor XCEL device (ATCOR, Naperville, IL, USA) ²⁷ after at least 10 min of rest in the supine position in the morning hours (9.00 a.m. to 11.00 a.m.). Pulse wave on the right carotid artery was detected using an applanation tonometer with an MM3 module, while on the right femoral artery through pulsations registered by a cuff placed around the thigh. The direct distance between the carotid artery and the top edge of the femoral cuff was measured.

The automatically‐corrected distance between the pulse wave detection points (the carotid and femoral arteries) and the transit time were used to calculate PWV with the formula: PWV [m/s] = 0.8 × (direct distance [m]/transit time [s]). ²⁸ Three measurements were performed at 1‐min intervals; the result with the best reproducibility was selected for analysis.

Pulse wave analysis (PWA) was performed by detecting the pulse wave on the brachial artery using a cuff placed on the arm. ²⁹ Based on this data and a built‐in validated algorithm called the transfer function, BP and waveform obtained from the brachial artery were transformed into an aortic pulse waveform to determine the central arterial pressure.

2.5. Ambulatory arterial stiffness index

Based on ABPM data, AASI was generated for each study participant using the formula: AASI = 1–α, where α denotes the regression slope of DBP on SBP. ³⁰ The less elastic the arteries are, the higher the AASI, ranging from 0 to 1.

2.6. Linear regression analysis of BP parameters

Linear regression of BP parameters (LRBPP) based on 24‐h ABPM readings was performed for every patient and the hemodynamic phenotype was determined according to individual regression coefficients. The LRBPP classifies patients into one of three phenotypes: (1) harmonious (H), (2) diastolic dysfunctional (D), and (3) systolic dysfunctional (S), as well as phenotype subtypes (classes) as described by Voitikova and Khursa. ²¹ , ³¹

Systemic circulation is generated by the pressure difference between the aorta and the right atrium. The ascending aorta wall distension in response to blood ejected from the left ventricle (LV) generates a pulse wave, which is transmitted peripherally along the vascular bed. Blood flow velocity depends on the elastic strain exerted by the blood wave on the arteries and the site of the pulse wave reflection (resulting in its return toward the aorta during the systole or the diastole). During each heartbeat, arterial BP varies between SBP and DBP, and their difference, that is, PP, correlates with blood ejection from the LV and arterial stiffness. According to the concept of the hemodynamic phenotype determination by the LRBPP model, a series of an individual's BP values, obtained in a time interval, represents the process of blood circulation (hemodynamics) as a functional system of interacting parameters, DBP and PP. PP is the result of the interaction between the contractile function of the heart and the “peripheral component,” vessels. PP, as a pressure gradient, ensures the movement of blood through the vessels and therefore it is an important differentiated characteristic of hemodynamics (PP = SBP–DBP). PP is represented by several individual BP parameters in the functional hemodynamic system; thus, PP is a parameter in the linear regression with SBP and DBP values. ³¹

Each SBP value corresponds to a probability distribution of PP, therefore BP can be represented mathematically by a regression function of SBP on PP. ¹⁸ Based on this assumption, two regression coefficients can be calculated according to the following formulas: SBP = Q+aPP, and DBP = Q+(a−1)PP [DBP = SBP–PP → DBP = Q+aPP–PP → DBP = Q+(a−1)PP]. ²¹ , ²³

Angular coefficients (slope) a and (a−1) show the proportion of change in the result (i.e., SBP and DBP, respectively) when the argument (PP) changes on unit and constitute a variable component of BP since it specifies the proportion between the LV and the “peripheral component”, in maintaining circulation. Slope a characterizes first of all the heart component, (a−1) – the “peripheral” component (primarily, vessels); intercept Q is the static component of the blood flow (at PP = 0), that is, theoretically corresponds to the mean hemodynamic pressure (mean pressure, MP). Assuming Q = MP, it is obvious that the known normal physiological ratio SBP > MP(Q) > DBP is possible only at 0 < a < 1 (harmonious phenotype H, Figure 1A). However, other phenotypes are also observed among people with varying frequencies depending on health status and age. Thus, phenotype H was observed only in 65%−75% of healthy normotensive young people. ²² , ³¹ , ³² If a < 0 or a > 1, unfavorable, dysfunctional hemodynamic phenotypes are recognized. In the case of a > 1, then SBP > DBP > Q; it is the D (diastolic dysfunctional) phenotype (Figure 1B), which indicates that the LV maintains circulation to a greater extent than physiologically: the pulse wave predominantly depends on the blood ejection from the (possibly hypertrophied) LV in light of the peripheral resistance due to arterial stiffness and/or insufficient participation of the peripheral component (the “peripheral heart”). If a < 0, then Q > SBP > DBP; the S (systolic dysfunctional) phenotype (Figure 1C) is recognized, which indicates an abnormally high role of the peripheral component in maintaining the circulation, we assume, presumably due to increased blood volume or in case of venous insufficiency. In clinical practice, many people have dysfunctional phenotypes, in particular, phenotype D present in young patients, especially with insufficient physical activity (∼8%−20%); in hypertensive patients (≥40%). The prevalence of phenotype D increases with age in the general population. Phenotype S is very rare in general population (∼1.5%−3.5%), more common in young athletes (∼5%−6%) and very rare in older people and hypertensive patients (∼0.5%−1.5%) ³¹ , ³²

FIGURE 1.

Graphical presentation of blood pressure parameters including type of hemodynamic functional status, that is, harmonious phenotype (A), diastolic dysfunctional phenotype (B) and systolic dysfunctional phenotype (C). The figure represents the work of Raissa Khursa. a, regression coefficient; Q regression coefficient Dotted line – Q value, blue line – DBP; red line – SBP. DBP, diastolic blood pressure; PP, pulse pressure; SBP, systolic blood pressure.

In patients with both dysfunctional phenotypes, linear regression suggests abnormal relations between SBP, Q and DBP, which is impossible physiologically. These abnormal conditions were investigated by a method of testing surrogate time series ³³ and as it was established, the BP series of patients with phenotype H is weakly nonlinear and the linear correlation between SBP and DBP in this phenotype is low. In patients with both dysfunctional phenotypes (D and S), the BP series were highly nonlinear and the correlation between SBP and DBP was strong. Therefore, in both dysfunctional phenotypes, Q does not correspond to the actual MP. ³¹ , ³⁴ The authors hypothesize that in the phenotype H, SBP, and DBP are formed mainly by hemodynamic mechanisms: by the heart's automatism, its ability to eject blood into systole with a force corresponding to the volume of blood entering it in diastole and the peripheral resistance of the vessels; and by the vasculature–basal tone and the ability to stretch according to the volume of blood entering in systole and compress, shifting it. We hypothesize that the close direct relationship between SBP and DBP in dysfunctional phenotypes may be a sign of interference by higher levels of regulation in the process of blood propagation, in addition to hemodynamic mechanisms. ³¹ Thus, determining the hemodynamic phenotype by LRBPP can separate patients with different characteristics of the blood circulation process. LRBPP phenotypes are stable individual characteristics, independent of BP, but may alter with treatment, lifestyle modification, and ageing. ²² , ²³

Taking into account both regression coefficients (a and Q) and using a diagnostic map (nomogram), the following phenotype subtypes (or hemodynamic classes) can be identified: normotensive, hypotensive, and hypertensive. The nomogram was created by application of Data Mining–discriminative algorithm Support Vector Machine (SVM); which allows to diagnose hemodynamic classes (for 24‐h ABPM period): dysfunctional diastolic (D2), harmonious (H2), dysfunctional systolic (S2)‐in normotensive patients; dysfunctional diastolic (D1), harmonious (H1), dysfunctional systolic (S1)‐in hypotensive patients; dysfunctional diastolic (D3), harmonious (H3), dysfunctional systolic (S3)‐in hypertensive ones. These are determined by applying of the patient's regression coefficients on the nomogram. The hemodynamic class is determined by identifying the point of intersection of coefficients a and Q on the nomogram relative to the separating borderlines (Figure 2). ²¹ , ³⁴

FIGURE 2.

The nomogram based on linear regression of blood pressure parameters (LRBPP) coefficients for hemodynamic classes defining (24‐h ambulatory blood pressure monitoring): D1, D2, D3; H1, H2, H3 and S1, S2, S3 – respectively, diastolic dysfunctional (D), harmonious (H) and systolic dysfunctional (S) phenotypes: hypotensive (1), normotensive (2) and hypertensive (3). Figure adapted from [21].

2.7. Statistical analysis

Results were expressed as numbers, number (percentage, %), mean ± standard deviation (SD) for parametric and median (interquartile range, IQR) for nonparametric data. Shapiro–Wilk test was used to assess the normality of data distribution. Pearson's and Spearman's correlations were calculated according to data distribution. Differences in binary variables were verified with Fisher's exact test. Differences in continuous variables between groups were evaluated using the Student's t‐test or Mann–Whitney U‐test, depending on data distribution. When three groups were compared, ANOVA or Kruskal–Wallis test was used. Outliers were detected with the ROUT method with Q set at 5%. Statistical analysis was performed with GraphPad Prism 9.4 software with one exception: RStudio 2023.06.0 was used to compare differences in binary variables between three groups with Fisher's exact test with a Benjamini‐Hochberg false discovery rate method correction for multiple testing. A value of p < .05 was considered as significant.

3. RESULTS

3.1. Patients’ clinical characteristics

Clinical and hemodynamic characteristics of the study patients are presented in Table 1. Patients were aged 60 ± 11 years, 21 (40%) were males. Age (60 ± 9 vs. 60 ± 12 years, p = .84) and BMI (30 ± 4 and 29 ± 6 kg/m², p = .68) were comparable between men and women (Table S1).

TABLE 1.

Clinical and hemodynamic parameters in a study group and subgroups according to chronic kidney disease (CKD) stage and linear regression of blood pressure parameters (LRBPP) hemodynamic phenotype.

Patients	All	G3a	G3b	G4	p	Phenotype H	Phenotype D	p
n	52	28	16	8	–	26	26	–
LRBPP phenotype (n/n)/CKD stage (n/n/n)	–	Phenotype H/D				Stage G3a / G3b / G4
		14/14	7/9	5/3	n.s.	14/7/5	14/9/3	n.s.
Age (years)	60 ± 11	64 (54–68)	64 (55–68)	59 (51–66)	<.001	57 ± 11	63 ± 10	.04
M/F (n/n)	21/31	21/7	0/16	0/8	<.001	14/12	7/19	.04
BMI	30 ± 6	30 ± 5	30 ± 6	28 ± 8	.27	30 ± 6	29 ± 5	.89
Smokers (n[%])	4 (8%)	0	2 (12%)	2 (25%)	n.s.	0	4 (15%)	.11
eGFR (mL/min./1.73m2)	44 ± 12	54 ± 5	38 ± 4	24 ± 3	.001	44 ± 12	45 ± 12	.8
HT (n[%])	45 (86%)	25 (89%)	14 (87%)	6 (75%)	.47	21 (81%)	24 (92%)	.42
HT medications (n) a	3 (2–4)	2.5 (1–4)	3 (1.25–3.75)	4 (1.75–5)	.29	3 (2–4)	3 (2–4)	.97
Controlled HT (n[%]) a	33 (73%)	17 (61%)	11 (69%)	5 (63%)	.1	19 (73%)	14 (54%)	.25
Diabetes (n[%])	7 (13%)	5 (18%)	1 (6%)	1 (13%)	.74	2 (8%)	5 (19%)	.42
mrEF HF (n[%])	5 (10%)	4 (14%)	0 (0%)	1 (13%)	.28	3 (12%)	2 (8%)	.99
Office SBP (mm Hg)	130 (124–141)	127 (115–132)	127 (114–143)	129 (123–134)	.88	128 (122–130)	137 (130–150)	.001
Office DBP (mm Hg)	77 ± 9	79 (73–81)	76 (68–87)	80 (76–86)	.99	74 ± 8	77 ± 10	.93
24‐h SBP (mm Hg)	124 (114–136)	124 (117–136)	125 (116–144)	120 (112–145)	.94	118 (111–128)	129 (122–146)	.001
24‐h SBP SD (mm Hg)	13.7 ± 3.5	13.6 ± 3.4	13.9 ± 3.5	13.8 ± 4.3	.42	11.7 ± 2.3	15.7 ± 3.4	<.001
24‐h DBP (mm Hg)	71 (66–80)	73 (65–79)	71 (68–83)	75 (64–86)	.9	73 (68–79)	70 (65–85)	.61
24‐h DBP SD (mm Hg)	10.4 ± 2.4	10.1 ± 2.5	10.5 ± 1.9	10.4 ± 2.3	.16	10.4 ± 2.4	10.2 ± 2.2	.81
24‐h PP (mm Hg)	53 ± 13	53 ± 13	54 ± 14	55 ± 12	.98	47 (40–50)	57 (51–65)	.001
24‐h HR (bpm)	71 ± 8	73 ± 9	68 ± 7	74 ± 9	.17	74 ± 9	70 ± 8	.08
Daytime SBP (mm Hg)	128 (118–142)	127 (119–140)	12 (11–15)	119 (112; 150)	.82	121 (112–131)	132 (127–149)	<.001
Daytime DBP (mm Hg)	74 (68–79)	76 (68–81)	9 (8–10)	75 (67–84)	.97	75 (70–81)	74 (66–88)	.84
Nighttime SBP (mm Hg)	114 (108–127)	114 (108–127)	116 (109–131)	111 (108–139)	.89	110 (106–119)	116 (112–134)	.04
Nighttime DBP (mm Hg)	63 (59–72)	65 (59–72)	64 (59–71)	66 (58–72)	.99	71 (66–82)	61 (58–78)	.31
Nondippers (n[%]) a	27 (60%)	11 (79%)	11 (69%)	5 (63%)	.059	17	10	.09
AASI	0.36 ± 0.14	0.35 ± 0.11	0.32 ± 0.15	0.43 ± 0.19	.09	0.31 ± 0.1	0.40 ± 0.1	.02
PWV (m/s)	9 (7.7–10.9)	9 (8–10.6)	9.8 (7.5–12.1)	7.7 (6.7–9.6)	.19	8.2 (7.3–10.3)	9.7 (8.3–10.9)	.02
Central SBP (mm Hg)	122 ± 17	123 ± 15	122 ± 20	122 ± 18	.97	115 ± 12	129 ± 18	.01
Central DBP (mm Hg)	76 ± 8	75 ± 3	75 ± 11	81 ± 9	.22	75 ± 7	76 ± 10	.75
Central PP (mm Hg)	43 (33–56)	43 (33–61)	47 (35–55)	38 (33–43)	.51	36 (32–43)	48 (40–67)	.001
Q	81 ± 22	82 ± 23	82 ± 21	75 ± 26	.72	87 (75–98)	69 (55–84)	.002
A	1.04 ± 0.43	1.01 ± 0.47	0.96 ± 0.43	1.31 ± 0.21	.15	0.75(0.53–0.93)	1.33(1.21–1.55)	.001
ACE‐I/ARB (n[%]) a	40 (77%)	25 (89%)	11 (69%)	4 (50%)	.03	18(69%)	22(85%)	.32
ß‐blocker (n[%])	29 (56%)	16 (57%)	10 (63%)	3 (38%)	.55	14(54%)	15(58%)	.99
DHP CCB (n[%]) a	29 (56%)	14 (50%)	9 (56%)	3 (38)	.69	12 (46%)	17 (65%)	.26
Thiazide (n[%])	20 (38%)	11 (39%)	7 (44%)	2 (29%)	.74	7 (27%)	13 (50%)	.15
Loop diuretic (n[%])	8 (15%)	4 (14%)	2 (13%)	2 (29%)	.75	5 (19%)	3 (12%)	.70
MRA (n[%]) a	3 (6%)	2 (7%)	1 (6%)	0 (0%)	1	3 (12%)	0 (0%)	.23
α‐blocker (n[%]) a	14 (27%)	9 (32%)	2 (13%)	3 (38%)	.26	7 (27%)	7 (27%)	1

Abbreviations: α‐blocker, alpha‐blocker; AASI, ambulatory arterial stiffness index; ACE‐I, angiotensin converting enzyme inhibitor; ARB, angiotensin II receptor blocker; ß‐blocker, beta‐blocker; bpm, beats per minute; DHP CCB, dihydropyridine calcium channel blocker; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; HR, heart rate; MRA, mineralocorticoid receptor antagonist; mrEF HF, mildly reduced ejection fraction heart failure; PP, pulse pressure; PWV, pulse wave velocity; SBP, systolic blood pressure.

^a Only patients with hypertension were considered

Mean eGFR was 44 ± 12 mL/min/1.73 m². Twenty‐eight (54%) patients were classified to stage G3a CKD, 16 (31%) to G3b and 8 (15%) to G4.

HT was previously diagnosed in 45 (86%) patients. Mean office SBP/DBP was 133 ± 17/77 ± 9 mm Hg and heart rate was 68 ± 11 bpm. Among patients with HT, median number of anti‐hypertensive drugs was 3 (2–4) and there were no significant differences in the use of BP‐lowering drug classes between phenotype H and D patients (Table 1).

Seven (13%) patients had type 2 diabetes. Five patients had chronic heart failure with mildly reduced LV ejection fraction (HFmrEF), which ranged between 40% and 55%. Twenty‐six (50%) patients were treated with statins. Four patients (8%) were active smokers.

3.2. LRBPP‐derived hemodynamic phenotypes, clinical, and ABPM parameters

Median 24‐h SBP/DBP was 124 (114–136)/71 (66–80) mm Hg, daytime SBP/DBP 128 (118–142)/74 (68–79) mm Hg, nighttime SBP/DBP 114 (108–127)/63 (59–72) mm Hg (Table 1).

Nineteen (37%) patients had uncontrolled HT (mean 24‐h SBP/DBP values ≥130/80 mm Hg and/or mean daytime ≥135/85 mm Hg and/or mean nighttime ≥120/70 mm Hg).

There was a comparable number of patients classified to phenotype H (n = 26) and D (n = 26). Five out of seven CKD patients who were normotensive were classified to phenotype H.

In the study, 16 (31%) patients demonstrated H2 (harmonious normotensive), 10 (19%) H3 (harmonious hypertensive), 6 (12%) D2 (diastolic dysfunctional normotensive), and 20 (38%) D3 (diastolic dysfunctional hypertensive) hemodynamic phenotype class. There were no patients categorized to hypotensive (H1 and D1) classes.

Phenotype D patients were significantly older compared to phenotype H (63 ± 10 vs. 57 ± 11 years, p = .04). There was a similar number of men (n = 14) and women (n = 12) in phenotype H, while in phenotype D there were more females (n = 19) than males (n = 7). There were no significant differences in eGFR (45 ± 12 vs. 44 ± 12 mL/min/1.73 m², p = .8) between patients with D and H phenotype. All smokers (n = 4) were classified to phenotype D.

Patients with phenotype D compared to those with phenotype H had significantly higher median 24‐h SBP (129 [122–146] vs. 118 [111–128] mm Hg) and PP (57 [51–65] vs. 47 [40–50] mm Hg) (p = .001 for all), whereas DBP was comparable (73 [68–79] vs. 70 [65–85] mm Hg, p = .61). 24‐h systolic BPV was lower in patients with phenotype H than phenotype D (11.7 ± 2.3 vs. 15.7 ± 3.4 mm Hg, p < .001). Despite a higher median of 24‐h SBP, fewer non‐dippers were observed among patients with phenotype D (10 vs. 17, p = .09). The 24‐h HR did not differ significantly between profile H and D patients (74 ± 9 vs. 70 ± 8 bpm, p = .08).

3.3. LRBPP‐derived hemodynamic phenotypes, PWV, PWA, and AASI

Median PWV of all patients was 9 (7.7−10.9) m/s. PWV was higher in phenotype D patients compared to phenotype H (9.7 [8.3–10.9] vs. 8.2 [7.3–10.3] m/s, p = .02). The relationship between variables and PWV is summarized in Table 2. PWV was associated with age (r = 0.43, p = .002), however, this association remained significant in phenotype H (r = 0.51, p = .008) but not in phenotype D (r = 0.19, p = .35) patients. There was no significant correlation between PWV and eGFR. PWV was related to 24‐h SBP, systolic BPV, 24‐h diastolic BPV and PP (Table 2).

TABLE 2.

The correlation between pulse wave velocity (PWV) and study variables.

PWV	All patients		Phenotype H		Phenotype D
	r	p	r	p	r	p
Age	0.43	.002	0.51	.008	–	.35
eGFR	0.25	.08	0.19	.36	0.36	.07
24‐h SBP	0.49	.0001	0.60	.001	–	.24
24‐h SBP SD	0.39	.005	0.45	.02	–	.55
24‐h DBP SD	0.32	.02	0.48	.01	–	.35
PP	0.38	.05	0.49	.01	–	.34
Nighttime SBP	0.39	.04	–	.09	–	.05
Nighttime PP	0.56	<.0001	0.52	.007	0.57	.003
AASI	–	.16	–	.22	−0.43	.03
cSBP	0.58	<.001	0.54	.005	0.51	.009
cDBP	–	.38	–	.31	–	.77
cPP	0.69	<.0001	0.63	.0001	0.57	.003

Note: Correlation coefficient is given or lack of significance is marked with a minus sign. The analysis was performed after excluding outliers: two for PWV, two 24‐h SBP. There were no statistically significant correlations between PWV and: BMI, eGFR, 24‐h DBP, nighttime SBP SD, nighttime DBP and cDBP.

Abbreviations: cDBP, central diastolic blood pressure; cPP, central pulse pressure; cSBP, central systolic blood pressure.

Univariate regression analysis showed that PWV was associated with age, eGFR, and 24‐h SBP, while the statistical significance of LRBPP hemodynamic phenotype was 0.21 (p was even higher for BMI, sex, and smoking status). The multivariate regression analysis showed the influence of age (beta = 0.112, p < .01) and 24‐h SBP (beta 0.042, p = .01) on PWV value. Neither eGFR nor hemodynamic phenotype were shown to predict PWV (Table S2).

Mean central SBP was 122 ± 17 mm Hg and central DBP was 76 ± 8 mm Hg in all patients. Mean central SBP (115 ± 12 vs. 129 ± 18 mm Hg, p = .01) and central PP (36 [32; 43] mm Hg vs. 48 [40; 67] mm Hg, p = .001) were lower in patients with phenotype H than D (Table 1). Mean central DBP was comparable (75 ± 7 vs. 76 ± 10 mm Hg, p = .75) between phenotype H and D patients.

Mean AASI was 0.36 ± 0.14, and was lower in patients with phenotype H than D (0.31 ± 0.1 vs. 0.40 ± 0.1, p = .02). AASI was inversely associated with eGFR (r = −0.23, p = .05) (Figure 3A). This correlation was present in phenotype D (r = −0.43, p = .03, Figure 3B) but not in phenotype H (r = −0.18, p = .37, Figure 3C) patients. AASI was inversely related to PWV in patients with phenotype D (r = −0.43, p = .03) but not with phenotype H (Table 2).

FIGURE 3.

Correlation between estimated glomerular filtration rate (eGFR) and ambulatory arterial stiffness index (AASI) in all patients (A), phenotype D (B), and phenotype H (C) patients.

4. DISCUSSION

The present study found that 50% of patients in stage G3‐G4 CKD demonstrate dysfunctional hemodynamic phenotype D, associated with elevated SBP, increased systolic BPV, and higher arterial stiffness indicated by PWV and AASI.

Previously, data from 50 newly diagnosed patients with HT, 121 healthy normotensives, and 43 patients with acute hypotension during intensive care unit hospitalization were analyzed by the LRBPP. ²¹ , ²² Individuals at risk of developing HT and acute hypotension were identified, corresponding to H normotensive and D hypotensive classes. Among patients with HT, both H and D (especially, in long‐standing disease) phenotypes were observed. ²¹ , ²²

In another study, young healthy patients (n = 120, age 24.5 ± 0.3) were compared to young patients with newly diagnosed HT (n = 45, age 29.1 ± 0.7). ³² The D phenotype was more common among those with HT (51% vs. 20%) and associated with significantly higher PWV, this parameter was comparable between controls and hypertensive patients. ³² A further report revealed that among 33 hypertensive patients with OSA, the best hemodynamic response to continuous positive airway pressure therapy (at least 5% nighttime BP dipping) was associated with the class H3 on LRBPP, and 18.2% of patients were classified as D1‐class, which is associated with high risk of acute hypotensive episodes. ²³

Аmong 175 hypertensive patients aged 60 (52–70), phenotype D was associated with a lower rate of HT control and quality of life. ³⁵

PWV in our patients was comparable to that recorded both in a small sample (n = 45) of stage G3 and G4 CKD patients (median 8.6 [6.9–10.5] m/s) by Zanoli and coworkers ³⁶ and CKD cohort of approximately 2.500 G3–G4 CKD patients (eGFR 40 ± 15 mL/min/1.73 m², age 60 ± 11 years) whose mean aortic PWV was 9.5 ± 3 m/s. ³⁷ In both these studies and another CKD patient sample (n = 806), a negative correlation between PWV and eGFR was found. ³⁸ However, this association was not reported in two other studies with 95 and 150 CKD patients, whose PWV was higher (respectively, 11.3 ± 2.7 and 14.6 ± 3.8 m/s). ³⁹ , ⁴⁰ The lack of the PWV‐eGFR correlation in our patients probably results from the clinical characteristic of the group (i.e., exclusion of atherosclerotic disease and long‐lasting diabetes) to reduce the impact of these comorbidities on hemodynamic parameters.

Conflicting results also exist regarding the relationship between eGFR and AASI. While Wang and coworkers reported a negative correlation (n = 583, mean age 43 ± 16 years, mean eGFR 59 ± 48 mL/min/1.73 m²), ⁴¹ Boesby and coworkers found that higher CKD stage was associated with increased arterial stiffness assessed by AASI in 83 CKD stage G2‐G5 patients, but this association was insignificant after adjustment for age. ⁴² Although AASI was higher in hypertensive non‐diabetic patients with CKD (n = 30, mean eGFR of 35.3 ± 2.8 mL/min/1.73m²) when compared to patients with preserved kidney function, but it did not correlate with creatinine levels or eGFR. ⁴³ Given the high BP and HR variability in CKD patients, the usefulness of AASI determination in this patient cohort was questioned. ⁴³ Furthermore, AASI depends on the nocturnal BP fall to a significant extent and CKD patients are frequently non‐dippers. ⁴³ , ⁴⁴ Despite these facts, Boesby and coworkers reported a moderate yet acceptable reproducibility of AASI, suggesting that it may be a useful parameter in the evaluation of the CKD patients’ hemodynamic status. ⁴²

AASI has been proposed as a new marker of vascular stiffness, based on the data that for any given increase in arterial pressure, SBP, and DBP tend to increase in a parallel manner in a compliant artery, while in a stiff artery, the increase in SBP is accompanied by a lesser increase or by a decrease in DBP. ¹⁶ , ³⁰ Currently, the clinical and prognostic value of increased AASI (more than 0.5–in young people, more than 0.7–in older people) is widely accepted. AASI is an independent predictor of CV mortality and stroke and has been correlated with target organ damage. ¹⁶ , ³⁰ , ⁴⁵ AASI was found to be associated with chronic inflammation, CV events, ⁴⁶ and renal outcomes. ⁴⁷ However, the pathophysiological determinants of AASI remain unclear, given that AASI depends on the nocturnal decrease in BP, on the correlation between SBP and DBP, on patient age and is affected significantly by treatment. The low correlation coefficient between SBP and DBP leads to an artificial increase in AASI, which has been observed in non‐dipper patients. ⁴⁵ AASI may reflect not only arterial stiffness but also diurnal BP variability and/or autonomic nervous dysfunction. Moreover, AASI has a much stronger association with BP variability than with BP itself. ⁴⁸

Derivation of hemodynamic phenotypes by linear regression shows many similarities with AASI as both approaches are based on linear regression, but LRBPP a priori is based on the identification of individual characteristics of the hemodynamic process as the ratio of cardiac and peripheral components by analyzing the relationships between BP parameters. As a result, LRBPP allows the differentiation of H and D phenotypes and the D phenotype theoretically suggests a decrease in the peripheral component contribution (primarily represented by vasculature) due to an increase in the contribution of the cardiac component, which may be due to increased vascular stiffness, low physical activity or other causes. Our patients with phenotype D had higher AASI (0.40 ± 0.1) than those with phenotype H (0.31 ± 0.1, p = .02).

We found that dysfunctional phenotypes differ from the harmonic ones by the nonlinearity of the BP series and by the high correlation between SBP and DBP, while phenotype H is characterized by a low correlation between SBP and DBP. ³¹ , ³⁴ This is why in the present study of CKD patients significant differences in AASI between phenotypes D and H are not surprising. As previously shown by Schillaci and Parati, ⁴⁵ the low correlation between SBP and DBP present in non‐dipper patients tend to artificially increase. In our study group, there were more non‐dipper patients in phenotype H than in D (17 vs. 10 patients, respectively, p = .09) and patients with phenotype D had a higher AASI (0.40 ± 0.1) than patients with phenotype H (0.31 ± 0.1, p = .02). Considering that in our study PWV was significantly higher in phenotype D than in H, belonging to phenotype D likely indicates increased vascular stiffness to a greater extent than the AASI index. The fact that in our study PWV was significantly higher in phenotype D than phenotype H, confirms the same results, previously obtained in normotensive patients. ²² , ³² There is evidence to suggest that AASI is higher in stiffer arteries due to the nonlinear behavior of the arterial wall, ⁴² which is consistent with data about the nonlinearity of BP series in dysfunctional phenotypes. Possibly, the use of AASI based on linear analysis, in patients with different phenotypes and different correlations between SBP and DBP, leads to some bias of the result. We hypothesize that phenotype D could be considered as one of the signs of increased vascular stiffness. LRBPP phenotypes reflect the combined effects of various hemodynamic regulatory mechanisms, conveying their adaptation to a certain extent. Therefore, the LRBPP phenotype should be taken into account in conjunction with other clinical and hemodynamic parameters. The heterogeneous characteristics of our patients probably underlie the discrepancies in the results obtained when hemodynamic parameters were assessed using different methods.

All five of our patients with HFmrEF were classified as phenotype H, which is in line with the LRBPP assumptions: impaired myocardial contractility excludes the high LV output, characteristic of phenotype D. On the other hand, this fact may also indicate positive effects of pharmacotherapy: the stable clinical condition required for the study enrollment was confirmed by normal hemodynamics. In the Strategy for Preventing CV and Renal Events Based on Arterial Stiffness study, PWV decreased with angiotensin‐converting enzyme inhibitors or angiotensin receptor blockers and calcium channel blockers therapy. ⁴⁹ Another group of medications that was shown to mitigate arterial stiffness as evidenced by a decrease in PWV are sodium‐glucose cotransporter‐2 inhibitors. ⁵⁰

5. NOVELTY/PUBLIC IMPLICATIONS AND THE LIMITATIONS

The mathematical models of hemodynamics based on ABPM data can help to better understand the CV system and can be used in application to improve decision‐making, diagnosis or predict outcomes if studied in a large patent cohort. Achievements of medical technologies significantly increase the potential of ABPM, which is determined by the methodology of medical signals analysis. Phenotype determination by LRBPP is a new way to broaden the understanding of individual hemodynamic characteristics and find new markers of preclinical (latent) disorders that can be useful in clinical and preventive medicine, based on a dynamic series of BP parameters. This study aimed to investigate the prevalence of LRBPP phenotypes in patients with CKD (a condition not previously studied) and their association with some clinical and hemodynamic parameters, including vascular stiffness. Our findings suggest that further studies of LRBPP may be useful not only in patients with CKD but also in different populations and various pathologies.

There are some limitations of this study. First, the observation group was small, which limited the possibilities of correlation analysis, especially taking into account the division of the group into phenotypes. Secondly, the study was not longitudinal, which does not allow to conclude the prognostic significance of the LRBPPphenotype in patients with CKD, that is, whether the phenotype influences kidney function decline or death. Third, the pathophysiological mechanisms of LRBPP phenotypes have not been studied and are as unclear as AASI. Currently, the value of AASI (clinical, prognostic) is widely accepted, while clinical studies of phenotypes were carried out only in limited populations (in Belarus, Russia, and Poland) and the prognostic significance has not been investigated. We believe that AASI and LRBPPphenotypes are different in meaning hemodynamic characteristics, that are useful and complementary to each other, especially since their calculation does not require special expensive equipment.

6. CONCLUSIONS

The essence of the linear regression analysis of BP parameters with the hemodynamic phenotypes defining is to determine the relative contribution of the heart (the LV) and the “peripheral component” (primarily the vasculature) in maintaining blood circulation. This study showed that among stage G3‐G4 CKD patients without clinically overt CV disease, the D phenotype is associated with unfavorable indices of hemodynamics such as increased SBP, BPV, PWV, and AASI.

The LRBPP model may be useful in detecting latent abnormalities in patients with controlled BP. Determining the hemodynamic phenotype by LRBPP from 24‐h ABPM seems to be an easily available non‐invasive tool to potentially identify individuals predisposed to CV complications. Longitudinal studies are required to establish the clinical value of the LRBPP model.

AUTHOR CONTRIBUTIONS

Conception and study design, Katarzyna Cierpka‐Kmieć and Dagmara Hering; methodology, Katarzyna Cierpka‐Kmieć, Dagmara Hering, and Raissa Khursa; analysis and interpretation of data Katarzyna Cierpka‐Kmieć, Dagmara Hering, and Raissa Khursa; drafting the manuscript Katarzyna Cierpka‐Kmieć, Dagmara Hering, and Raissa Khursa; supervision, Dagmara Hering; funding acquisition, Dagmara Hering. All authors have read and approved the final version of the manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

Supporting information

Supporting information

Cierpka‐Kmieć K, Khursa R, Hering D. Hemodynamic phenotypes in chronic kidney disease patients based on linear regression of blood pressure parameters. J Clin Hypertens. 2024;26:1391–1401. 10.1111/jch.14880

DATA AVAILABILITY STATEMENT

The data presented in this study are available on request from the corresponding author.