Correlation of Renal Oxygenation with Renal Function in Chronic Kidney Disease: A Preliminary Prospective Study

Chronic kidney disease (CKD) remains a major public health problem with the increasing incidence and high mortality worldwide [1]. The early onset of CKD is reclusive, without obvious clinical symptoms, which brings difficulties to the early diagnosis of the disease. Irrespective of the initial insult, renal fibrosis is the main pathological basis for the progression of CKD to end-stage renal disease [2]. With the in-depth studies on the pathogenesis of renal fibrosis, chronic hypoxia of renal parenchyma plays an important role in promoting CKD progression and fibrosis [3, 4]. Although almost 20–25% of the whole cardiac output was allocated to the kidney, the organ is still prone to hypoxia due to the diffusional shunting of O₂ between renal arteries and veins in physiological state [5]. In addition to innate low oxygen tension in the kidney, pathological conditions across the progression of CKD, such as inflammation, anemia, and increased proteinuria, also lead to severe hypoxia in the kidney [6]. Thus, renal oxygenation needs precise dynamic control to avoid hypoxic injury, which requires effective detection methods of oxygen tension.

Blood oxygenation level-dependent magnetic resonance imaging (BOLD-MRI), using deoxyhemoglobin paramagnetic characteristics, can noninvasively evaluate renal tissue oxygenation by an intrinsic tissue MR parameter, i.e., the effective transverse relaxation time T2* (or the effective transverse relaxation rate R2*, R2* = 1/T2*) [7, 8]. A lower T2* value (or higher R2* value) refers to greater concentration of deoxyhemoglobin, which means greater tissue hypoxia. BOLD-MRI has been applied in several studies as a promising technique and has shown high diagnostic value for measuring renal oxygenation. Pruijm et al. [9] found that low cortical oxygenation evaluated by BOLD-MRI could be an independent predictor for renal function decline. Furthermore, different analyzed techniques of BOLD-MRI have been used in previous studies and confirmed the reliability of T2* or R2* values [10, 11]. However, despite the advantages mentioned above, there were still debates about the validity of BOLD-MRI [12], which requires larger and multicenter prospective studies to be demonstrated.

Our previous study has demonstrated the feasibility and clinical value of BOLD-MRI for evaluating renal oxygenation in patients with CKD [13]. Hence, the purpose of this research was to observe the renal oxygenation in patients with CKD by BOLD-MRI and to further study the correlation of local oxygenation with renal function in different stages of CKD.

This prospective observational study was performed with approval from the Ethics Committee of the Shuguang Hospital affiliated to Shanghai University of Traditional Chinese Medicine (Identification number: 2019-703-58-01), and all the participants signed subject consent. Data on demographics, clinical characteristics, and biochemical assessments were collected at the time of enrollment.

The study initially recruited 10 healthy volunteers (HVs) and 106 adult patients with stages 1–5 CKD from May 2020 to November 2021 at the Department of Nephrology in the Shuguang Hospital. Diagnosis of CKD was based on the Kidney Disease Outcomes Quality Initiative (K/DOQI) proposed by the National Kidney Foundation [14], and the baseline estimated glomerular filtration rate (eGFR) was calculated using CKD Epidemiology Collaboration (CKD-EPI) equations. All the patients with CKD according to their eGFR were allocated to 2 groups as follows: CKD 1–3 group, which included patients with CKD stages 1–3 (eGFR ≥30 mL/min/1.73 m²), and CKD 4–5 group, which included patients with CKD stages 4–5 (eGFR ≤29 mL/min/1.73 m²). 10 HVs were assigned to a control group (HV group). Inclusion criteria for the participants also require: aged >18 years, with stable vital signs, and no contraindications for MRI. Patients with diabetes mellitus, atherosclerotic nephropathy, renal arterial stenosis, autosomal dominant polycystic kidney disease or presence of multiple acquired cysts, or on dialysis, or who had a contraindication to MRI examination, and/or who declined to participate in this study were excluded. All the participants maintained a low dietary sodium intake within 3 days before MRI scanning and were asked to fast for 4–6 h before the MRI examination. The details for excluded patients were shown in Figure 1.

All BOLD-MRI measurements were performed on a 3.0 T magnetic resonance scanner (MAGNETOM Skyra; Siemens Healthcare, Erlangen, Germany) as our previous clinical study described [13]. All participants underwent routine T1-weighted imaging, T2-weighted imaging, and BOLD imaging for each kidney using a coronal multi-echo (7 echoes) gradient echo sequence with echo times of 2.46, 4.92, 7.38, 9.84, 12.30, 14.76, and 17.22 ms; a repetition time of 232 ms; a slice thickness of 3.5 mm; a flip angle of 60°; a bandwidth of 470 Hz/Px; a field of view of 380 mm; and a 168 × 256 matrix.

T2* maps for each kidney were automatically generated in line immediately after the acquisitions, and cortical and medullary T2* values (COT2* and MET2*) were evaluated separately by two experienced radiologists using six regions of interest (ROIs) placed at the upper, middle, and lower areas of the central slice, respectively, avoiding vessels, renal sinuses, and susceptibility artifacts carefully (shown in Fig. 2a–f). COT2* and MET2* were averaged in bilateral kidneys.

Analyses were performed with SPSS 26.0 and GraphPad Prism 9.3. Normal distributions for continuous variables were evaluated by the Shapiro-Wilk test. All parameters were expressed as mean ± SD, the number with percentage, or median with interquartile range, as appropriate. Single factor analysis of variance (one-way ANOVA), Kruskal-Wallis H test, and Student’s t test were used as appropriate to evaluate the differences of the hemoglobin, SCr, BUN, Cys C, UA, eGFR, 24-h urinary protein (24-h Upr), COT2*, and MET2* values among HV, CKD 1–3, and CKD 4–5 groups. We analyzed the reproducibility of BOLD-MRI by intraclass correlation coefficient (ICC). A Spearman or Pearson correlation coefficient was used to test the relationships between BOLD-MRI parameters (COT2* and MET2* values) and clinical characteristics. Receiver operating characteristic (ROC) curve analysis was used to evaluate the efficiency (sensitivity and specificity) of COT2* and MET2* values for assessing renal oxygenation in patients with CKD. A two-tailed p value <0.05 was considered to be statistically significant.

In our previous study, we recruited HVs and patients with CKD stages 1–4 to assess inter-observer variability of the classical ROI-based method, and we found high reproducibility of both COT2* (ICC = 0.937) and MET2* (ICC = 0.910) [13]. In order to further assess the reproducibility of BOLD-MRI in patients with CKD stage 5, 10 patients with CKD stage 5 were recruited and analyzed by previous ROI-based method. The reproducibility of both COT2* (ICC = 0.903) and MET2* (ICC = 0.892) was still high in patients with CKD stage 5.

97 patients with CKD and 10 HVs were recruited for the study. Details of baseline demographics and clinical characteristics of the patients are shown in Table 1. Of the 97 patients, 10 (10.3%) had CKD stage 1, 20 (20.6%) had CKD stage 2, 27 (27.8%) had CKD stage 3, 30 (30.9%) had CKD stage 4, and 10 (10.3%) had CKD stage 5. 65 patients had chronic glomerulonephritis, 11 had hypertensive nephropathy, 10 had other causes of CKD, and 11 had CKD of unknown etiology. There were significant differences in the hemoglobin, SCr, BUN, Cys C, eGFR, COT2*, MET2*, and 24-h Upr values among HV, CKD 1–3, and CKD 4–5 groups (p < 0.001).

COT2* values were significantly higher than MET2* values in all participants, including HV, CKD 1–3, and CKD 4–5 groups (t = 38.722, p < 0.0001). There were significant differences among 3 groups of COT2* values (“HV group 70.50 ± 7.48 ms” vs. “CKD 1–3 group 54.69 ± 5.37 ms” vs. “CKD 4–5 group 39.14 ± 6.32 ms”; F = 143.928, p < 0.0001) and MET2* values (“HV group 40.90 ± 4.18 ms” vs. “CKD 1–3 group 30.50 ± 6.34 ms” vs. “CKD 4–5 group 18.58 ± 3.56 ms”; F = 97.776, p < 0.0001) (shown in Table 2; Fig. 3).

Pearson or Spearman correlations between the clinical characteristics and BOLD-MRI parameters (COT2* and MET2* values) were presented in Figure 4a. The annotations for absolute value of correlation coefficient are as follows: 0.8–1.0, very strong correlation; 0.6–0.8, strong correlation; 0.4–0.6, moderate correlation; 0.2–0.4, weak correlation; 0.0–0.2, very weak or no correlation. The COT2* and MET2* values were significantly positively correlated with the eGFR and hemoglobin, and COT2* and eGFR showed the highest coefficient (r_s = 0.8858, p < 0.0001), while there was no correlation between the COT2* and MET2* values and UA value (p > 0.05). The 24-h Upr value showed weak correlation with the COT2* value (r_s = −0.2301, p = 0.0265) and no correlation with the MET2* value (p > 0.05). However, urinary microproteins, including urinary alpha1-microglobulin, urinary beta2-microglobulin, and urinary retinol-binding protein, showed strong correlation with COT2* and MET2* values (shown in Fig. 4b). In order to investigate whether patients with different stages of CKD show different sensitivity to renal hypoxia, we analyzed correlations of eGFR and COT2* and MET2* in CKD 1–3 and CKD 4–5 groups, respectively, and found CKD 1–3 group reflected higher coefficient (COT2*: r = 0.7161, p < 0.0001; MET2*: r = 0.4713, p = 0.0002) of renal oxygenation and eGFR than CKD 4–5 group (p > 0.05) (shown in Fig. 5).

The diagnostic ability of radiomics can be evaluated by the ROC curve, where a bigger area under the ROC curve (AUC) indicates higher diagnostic accuracy [15]. The AUC was 0.9737 (95% CI: 0.9403–1.000) for COT2* between HV group and CKD 1–3 group and 0.9739 (95% CI: 0.9496–0.9982) between CKD 1–3 and CKD 4–5 groups (shown in Fig. 6a). The AUC was 0.9132 (95% CI: 0.8411–0.9852) for MET2* between HV and CKD 1–3 groups and 0.9557 (95% CI: 0.9214–0.9900) between CKD 1–3 and CKD 4–5 groups (shown in Fig. 6b), suggesting a high efficiency of renal cortex and medulla oxygenation measured by BOLD-MRI, in discriminating renal function among healthy people, patients with CKD stages 1–3, and stages 4–5. The Youden Index (J), a summary statistic of the ROC curve, defines the potential effectiveness of a test method. Generally, the maximal Youden Index determines the optimal diagnosis cut-point. In this study, we analyzed that the optimal cut-point was determined to be “< 61.17 ms” for COT2* value and “< 35.00 ms” for MET2* value between healthy people and patients with CKD stages 1–3. The optimal cut-point between patients with CKD stages 1–3 and stages 4–5 was determined to be “< 47.34 ms” for COT2* value and “< 25.09 ms” for MET2* value.

Based on above optimal cut-points, we calculated that COT2* value had a sensitivity of 91.23% and specificity of 100.00%, while MET2* value had a sensitivity of 77.19% and specificity of 100.00% for discrimination of healthy people and patients with CKD stages 1–3 (shown in Table 3). To discriminate patients with CKD stages 1–3 and stages 4–5, the COT2* value had a sensitivity of 90.00% and specificity of 92.98%, while MET2* value had a sensitivity of 97.50% and specificity of 80.70% (shown in Table 4). Collectively, these data indicate that BOLD-MRI parameters (COT2* and MET2* values) have high diagnostic ability for discriminating patients with different degrees of CKD by evaluation of renal oxygenation.

The major findings of the current study were as follows. (1) The classical ROI-based method of BOLD-MRI shows clinical feasibility and reliability for evaluating renal parenchyma oxygenation. (2) The COT2* values were significantly greater than MET2* values in both HVs and patients with CKD. Meanwhile, COT2* and MET2* values gradually declined as CKD progressed. (3) The COT2* and MET2* levels were positively associated with eGFR and hemoglobin but negatively associated with 24-h Upr (only associated with COT2*) and urinary microproteins. (4) The COT2* and MET2* values displayed good correlation with eGFR in patients with CKD stages 1–3, while they did not show significant correlation between these values in patients with CKD stages 4–5. (5) The COT2* and MET2* values were found to be potential diagnostic markers for detecting different stages of CKD.

BOLD-MRI provides a noninvasive method for evaluation of tissue oxygenation using deoxyhemoglobin as an endogenous marker, which can be applied to monitor the existence of renal hypoxia in patients with CKD. Comparing the analyzed techniques including the classical ROI method [16], concentric objects [10], and 12-layer CO method [11, 17], we chose ROI-based BOLD-MRI in our study. In order to assess reliability of the ROI method, we further demonstrated the high reproducibility of BOLD-MRI in stage 5 CKD patients based on our previous study [13].

The current study shows that the COT2* and MET2* values are lower in patients with CKD and gradually declined as CKD progressed, and the COT2* values will always be higher than MET2* values in both HVs and CKD patients. This finding is consistent with renal physiology and the theory that chronic renal hypoxia is prevalent in progressive renal disorders [3, 18]. Owing to the complex oxygen diffusion gradients and oxygen metabolic processes within kidneys, the oxygenation of renal medulla is lower than the cortex, which was proved by oxygen microsensors [19, 20]. With the progress of CKD, the presence of hypoxic inflammation, oxidative stress, and dysregulated angiogenesis also aggravates renal hypoxia [21, 23], which supports our finding that oxygenation of renal cortex and medulla is decreased gradually as renal failure progresses and shows the role of hypoxia in the development of CKD.

In the present study, the COT2* and MET2* values were found to be correlated with clinical indicators. The COT2* and MET2* values were positively associated with eGFR, hemoglobin, indicating that the severity of renal hypofunction correlates with the degree of renal hypoxia. This finding agrees with our previous study [13] but seems to conflict partly with the results obtained by Wang et al. [24] that cortex oxygenation did not show correlation with eGFR in patients with diabetic nephropathy. However, our study excluded the patients with diabetes mellitus in order to avoid the influence of blood glucose metabolism on renal oxygenation [25]. Simultaneously, the results of our non-diabetic study were consistent with another two studies, in which the researchers further revealed contradictory correlation of renal oxygenation with eGFR in patients with diabetic nephropathy (no significant correlation) and CKD without diabetes (significant correlation) [26, 27]. Therefore, the relationship between renal oxygenation and eGFR in patients with diabetic nephropathy, even just with diabetes mellitus, still needs to be confirmed by further research. For testing the relativity of eGFR and renal oxygenation in different stages of CKD, we compared two groups, i.e., CKD 1–3 and CKD 4–5 groups, respectively, and concluded the significant correlation of renal oxygenation with eGFR in CKD 1–3 group, but the relationship disappeared in CKD 4–5 group. The possible reason was that the kidneys have lost their compensative effect in patients with stage 4 or 5 CKD and have induced extreme renal hypoxia, which attenuates the difference of renal oxygenation between them. Thus, the findings demonstrate the reliability and efficiency of BOLD-MRI for evaluating renal function in CKD, especially in patients with CKD stages 1–3.

Our results found that 24-h Upr has weak negative correlation with the COT2* value as previous study reported [28] and focused on the negative correlation between urinary microproteins, including alpha1-microglobulin, beta2-microglobulin, retinol-binding protein, and renal T2* values. Generally, cortical hypoxia induces local microinflammation together with increased reactive oxygen species, which promotes oxidative stress and renal fibrogenesis, then aggravates proteinuria. While excess proteinuria, as a vicious cycle, increases oxygen consumption of reabsorption and aggravates renal hypoxia, it consequently leads to the progression of CKD [29]. Urinary microproteins were highly specific for renal tubular or glomerular disease and considered as effective biomarkers for early detection of renal abnormalities [30, 31]. Our finding provides piees of evidence to confirm the relativity between proteinuria and cortical hypoxia by BOLD-MRI and further observed the significant negative correlation between excretion of urinary microproteins and renal oxygenation of COT2* and MET2* values.

In addition, we obtained “cut-off” criteria for discriminating healthy people, patients with CKD stages 1–3, and CKD stages 4–5 by analysis of T2* values in three groups and found T2* values may be potential diagnostic markers for detecting different stages of CKD. Recently, Ping Liang et al. [32] also found similar diagnostic criteria (“cut-off” T2* values) for discrimination of patients with different degrees of renal injury. However, the recruited subjects were children under 18 years of age and with different analysis methods, which differs from our study.

The advantages of this study were that we recruited patients with stage 5 CKD and tested reproducibility of classical ROI methods among them, based on our previous study. Additionally, we had evaluated the diagnostic value of BOLD-MRI in predicting stages of CKD with horizontal comparison among groups. However, the current study still had several limitations. First, there was a relatively small sample size for the study. Second, the recruited patients were not restricted to take the medicines, which could be more consistent with clinical situations. However, medicines such as diuretics, hypertension may influence the renal oxygenation. Third, only 34% of the recruited patients have had kidney biopsies. Thus, we need further pathologic-classified study to eliminate the influence of underlying etiologic diversity on diagnostic reliability of BOLD-MRI.

In conclusion, we have demonstrated the reliability and efficiency of BOLD-MRI in evaluating renal oxygenation. This technique may be clinically used as a potential diagnostic method for predicting the severity of renal injury and the progress of CKD.

The study was approved by the Ethics Committee of the Shuguang Hospital affiliated to Shanghai University of Traditional Chinese Medicine (identification number: 2019-703-58-01), and written informed consent was obtained from all subjects prior to enrollment and participation. All methods were performed in accordance with the relevant guidelines and regulations.

The authors declare that there is no conflict of interest regarding the publication of this paper.

This study was supported by the National Natural Science Foundation of China (81973770), The Three Years Action Plan Project of Shanghai Accelerating Development of Traditional Chinese Medicine (ZY[2018–2020]-FWTX-7005), Key Disciplines Group Construction Project of Pudong Health Bureau of Shanghai (PWZxq2017-07), and Innovative Training Project of Shanghai University of Traditional Chinese Medicine (Y2020006).

Research idea; study design; and original draft, review, and editing of the manuscript: Yizeng Xu and Chen Wang; data curation and formal analysis: Yizeng Xu, Jing Yang, and Fang Lu; visualization and MR scanning: Fang Lu; and funding acquisition: Chen Wang and Chaoyang Ye. All authors read and approved the final manuscript.

The datasets generated and/or analyzed during the current study are not publicly available due to patient confidentiality but are available from the corresponding author on reasonable request.

Date of registration: December 24, 2018; registration number: ChiCTR1800020332.