Abstract
Objective
To define phenotypic subgroups of thrombotic microangiopathy (TMA) by integrating clinicopathologic data to identify patterns suggestive of pathogenic features and etiologies.
Patients and Methods
We retrospectively analyzed 283 patients with biopsy-confirmed renal TMA between January 1, 2010, and December 31, 2020. Sixty-four clinicopathologic variables were applied in consensus clustering.
Results
Three distinct clusters were identified. Cluster 1 (n=78), predominantly women (n=50, 64%), was enriched for drug-induced (n=23, 29%), lupus-associated (n=9, 12%), and post–bone marrow transplant TMA (n=8, 10%). It showed wide immune-complex staining (IgM, C3, κ, and λ) and subendothelial deposits (n=16, 21%), suggestive of immune-associated injury. Cluster 2 (n=92) included the oldest patients (54±18 years) and was dominated by drug-induced TMA (n=21, 23%; one-third owing to vascular endothelial growth factor inhibitor bevacizumab) and monoclonal gammopathy–associated TMA (n=10, 11%). It showed marked endothelial damage (n=65, 71%) and mesangiolysis (n=66, 72%) with little immune staining, suggesting a toxic-related injury. Cluster 3 (n=113) exhibited the worst outcomes, with 80% (24/30) progressing to renal failure, compared with 31% (12/39) in cluster 1 and 34% (16/47) in cluster 2. This cluster was characterized by the youngest age (45±14 years), mostly males (n=67, 59%), severe hypertension (n=75, 66%), and low estimated glomerular filtration rate (10 mL/min/1.73 m2; interquartile range, 7-16). Biopsies showed acute vascular lesions—wrinkling of capillary tufts (n=74, 65%), fibrin thrombi (n=67, 59%), mucoid intimal edema (n=85, 75%), and onion-skin–type hyperplasia (n=55, 49%)—with minimal deposits (n=1, 0.9%), suggesting a severe vascular lesion process.
Conclusion
Integrative clustering of clinical and histologic data in TMA identified 3 clinicopathologic phenotypes that may suggest underlying immune-, toxic-, or severe vascular lesion–related drivers. These findings support data-driven classification that may help prioritize diagnostic and therapeutic considerations.
Graphical abstract
Thrombotic microangiopathy (TMA) represents a histopathologic pattern of microvascular injury, encompassing multiple conditions and disorders that may either be hereditary or acquired. It is characterized by endothelial cell damage, intravascular thrombosis, and tissue ischemia.1,2 Clinically, it often presents with hemolytic anemia, microangiopathic thrombocytopenia, and end-organ damage. The kidneys are among the most frequently affected organs owing to their high perfusion—approximately 25% of the cardiac output—coupled with their extensive endothelial surface area and heightened vascular sensitivity, which together increase vulnerability to injury.3 Although the histologic features of TMA, such as mesangiolysis, capillary wall duplication, mucoid intimal edema, and fibrin thrombi, are well described, they are largely nonspecific and shared across a spectrum of etiologies.4,5
Over time, classification schemes have evolved to distinguish primary TMA syndromes, such as thrombotic thrombocytopenic purpura (TTP) and atypical hemolytic uremic syndrome (aHUS) from secondary forms, which include drug-induced TMA, infections like Shiga toxin–associated hemolytic uremic syndrome, malignant hypertension, autoimmune disease, malignancy, pregnancy, or transplantation.3, 4, 5, 6, 7 However, in clinical practice, the distinction between primary and secondary forms can be blurred. Patients may have overlapping features, multiple potential triggers, or ambiguous histology, which complicates the diagnostic workup and delays targeted therapy. Timely intervention is crucial, as specific therapies—such as plasma exchange in TTP or complement inhibition in aHUS—can be lifesaving.8,9 Yet, the efficacy of these interventions hinges on accurate identification of the underlying pathogenesis. Traditional diagnostic algorithms often rely on clinical suspicion, serologic testing, and exclusion, which may be inadequate in complex cases.
In this study, we applied an unsupervised machine learning approach to a large cohort of patients with biopsy-proven renal TMA to determine whether distinct clinicopathologic phenotypes could be identified at presentation, potentially reflecting divergent pathogenesis and etiologies.
Patients And Methods
Patient Selection
This retrospective study was approved by Mayo Clinic institutional review board (21-007513). We identified our cohort by first searching the Mayo Clinic Pathology Database for the diagnosis of TMA from January 1, 2010, to December 31, 2020. We excluded patients who did not have TMA as their primary diagnosis (Supplemental Appendix 1, available online at https://www.mcpdigitalhealth.org/) and those whose etiology remained undetermined despite review of the records (Figure 1).
Figure 1.
Flowchart of patient’s inclusion and exclusion. aTo determine the etiology of the TMA, 2 separate nephrologists (L.Z. and M.E.) reviewed each case independently. If there was disagreement, the cases were then reviewed together, and if there was mutual agreement on the etiology, the patient was included and if the 2 could not agree on the etiology, the case was excluded from the analysis. The study included patients who were considered internal cases who underwent biopsy and were treated at Mayo Clinic and external cases who were diagnosed at an outside institution and treated there but their biopsy was reviewed by a Mayo Clinic pathologist. TMA, thrombotic microangiopathy. bThe biopsies from the internal cases were reviewed again by a pathologist from Mayo Clinic (M.P.A.), and the results were included, however, for the external cases, owing to the absence of biopsy tissue at the time of the study, the data were gathered from review of the biopsy reports.
Data Collection
Data were collected at the time of kidney biopsy, including demographic characteristics, pertinent clinical and laboratory values, and TMA etiologies. Genetic testing data were collected if available. Kidney biopsy findings were recorded.
Machine Learning Analysis
Patients were stratified into clusters using consensus clustering, an unsupervised machine learning technique that groups individuals based on similarity across clinical features.10,11 The clustering analysis incorporated 64 variables (Supplemental Table 1, available online at https://www.mcpdigitalhealth.org/), encompassing demographic characteristics, clinical and laboratory parameters, and kidney biopsy findings. Because consensus clustering requires a complete feature matrix, missing values were imputed before analysis using the missForest algorithm. Resulting clusters were visualized using the uniform manifold approximation and projection, a nonlinear dimensionality reduction technique.12 The optimal number of clusters was determined through evaluation of the consensus matrix heatmap, mean consensus score, and cumulative distribution function plots.
Clinical Outcome Assessment
Development of renal failure (estimated glomerular filtration rate [eGFR] <15 mL/min/1.73 m2 or dialysis initiation), receipt of kidney transplantation, and all-cause mortality during follow-up were assessed. A composite outcome was defined as the occurrence of either renal failure or death.
Statistical Analyses
Continuous variables are presented as mean ± SD or median with interquartile range (IQR). Categorical variables are expressed as counts and percentages. Group comparisons were conducted using the Kruskal-Wallis test for continuous variables and the χ2 test for categorical variables to evaluate whether at least 1 group differs from the others. Kaplan–Meier and Cox proportional hazard model were used to assess clinical outcomes. Hazard ratio and 95% CI were reported. A P value less than .05 was considered statistically significant. Analyses were performed using JMP Pro version 14.3.0 (SAS Institute).
Results
Clinical Characteristics of the Cohort
In total, 283 patients with biopsy-proven TMA were included for the final analysis (Figure 1). The average age was 49±17 years, with 52% (n=146) women. Baseline clinical characteristics included an eGFR of 17 mL/min/1.73 m2 (IQR, 10-28), 24-hour proteinuria of 1.8 g (IQR, 0.7-3.2), systolic blood pressure of 183±44 mm Hg, and diastolic blood pressure of 110±32 mm Hg. Additional clinical variables are summarized in Table 1. Malignant hypertension was the most common association (n=98, 35%), followed by drug-induced TMA (n=54, 19%) and scleroderma renal crisis (SRC) (n=28, 10%). Less frequent causes are detailed in Table 2.
Table 1.
Patient Baseline Characteristics and Laboratory Values
| Variables | Total (N=283) | Cluster 1 (n=78) | Cluster 2 (n=92) | Cluster 3 (n=113) | P |
|---|---|---|---|---|---|
| Age at biopsy (y) | 49 ± 17 | 51 ± 19 | 54 ± 18 | 45 ± 14 | .001 |
| Male | 137 (48) | 28 (36) | 42 (46) | 67 (59) | .005 |
| History of CKD | 100 (35) | 15 (19) | 20 (22) | 65 (58) | .06 |
| History of hypertension | 160 (57) | 45 (58) | 50 (54) | 65 (58) | .06 |
| SBP at biopsy (mm Hg) | 183 ± 44 | 172 ± 37 | 168 ± 39 | 202 ± 44 | .001 |
| DBP at biopsy (mm Hg) | 110 ± 32 | 100 ± 27 | 99 ± 28 | 126 ± 32 | .001 |
| Hemoglobin (g/dL) | 9.7 ± 1.5 | 9.7 ± 1.5 | 10.1 ± 1.5 | 9.3 ± 1.4 | .001 |
| Platelets (×109/L) | 213 ± 119 | 203 ±± 96 | 260 ± 160 | 182 ± 77 | .001 |
| Schistocytes, present (%) | 176 (62) | 46 (59) | 49 (53) | 81 (72) | .02 |
| eGFR (mL/min/1.73 m2) | 17 (10-28) | 21 (15-34) | 25 (14-42) | 10 (7-16) | .001 |
| BUN (mg/dL) | 21 (21-34) | 21 (21-37) | 21 (21-36) | 21 (21-24) | .50 |
| Albumin (g/dL) | 3.5 ± 2.2 | 3.4 ± 0.6 | 3.4 ± 0.5 | 3.6 ± 3.4 | .80 |
| ESR (mm/h) | 39 ± 13 | 37 ± 15 | 37 ± 13 | 42 ± 11 | .01 |
| Complement 3 (mg/dL) | 92 (89-96) | 93 (88-96) | 94 (90-97) | 91 (89-93) | .62 |
| Complement 4 (mg/dL) | 26 (24-28) | 25 (23-28) | 25 (23-28) | 27 (26-29) | .95 |
| Proteinuria (g/24 h) | 1.8 (0.7-3.2) | 2.1 (1.1-4.0) | 1.7 (0.7-3.6) | 1.6 (0.5-2.6) | .10 |
| Hematuria, present (%) | 98 (35) | 33 (42) | 24 (26) | 41 (36) | .07 |
Values are n (%), mean ± SD, or median (IQR). P values denote overall differences across the 3 clusters. Significant P values indicate that at least 1 cluster differs from the others.
BUN, blood urine nitrogen; CKD, chronic kidney disease; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; ESR, erythrocyte sedimentation rate; SBP, systolic blood pressure.
Table 2.
Etiology of TMA in the Overall Cohort and in Each Cluster
| Etiology of TMA | Total (N=283) | Cluster 1 (n=78) | Cluster 2 (n=92) | Cluster 3 (n=113) | P |
|---|---|---|---|---|---|
| mHTNa | 98 (35) | 13 (17) | 10 (11) | 75 (66) | .0001 |
| Drug-induced | 54 (19) | 23 (29)b | 21 (23)c | 10 (9)d | .001 |
| SRC | 28 (10) | 7 (9) | 9 (10) | 12 (11) | .98 |
| aHUSa | 25 (9) | 6 (8) | 11 (12) | 8 (7) | .43 |
| APS | 19 (7) | 8 (10) | 7 (8) | 4 (4) | .17 |
| Post-BMT | 16 (6) | 8 (10) | 8 (9) | 0 (0) | .003 |
| MG | 14 (5) | 3 (4) | 10 (11) | 1 (0.9) | .004 |
| SLE | 13 (5) | 9 (12) | 3 (3) | 1 (0.9) | .002 |
| Pregnancy-related | 8 (3) | 1 (1) | 7 (8) | 0 (0) | — |
| MPD | 5 (2) | 0 (0) | 5 (5) | 0 (0) | — |
| TTP | 3 (1) | 0 (0) | 1 (1) | 2 (2) | — |
| Genetic testing, n | 34 | 11 | 12 | 11 | — |
| Normal | 10 (29) | 3 (27) | 3 (25) | 4 (36) | — |
| aHUS | 3 | 1 | 1 | 1 | — |
| Lupus | 2 | 2 | — | — | — |
| SRC | 1 | — | — | 1 | — |
| MG | 1 | — | 1 | — | — |
| APS | 1 | — | — | 1 | — |
| mHTN | 1 | — | — | 1 | — |
| Drug | 1 | — | 1 | — | — |
| Genetic mutation present | 24 (71) | 8 (73)e | 9 (75)f | 7 (64)g | — |
| aHUS | 12 (80) | 5 (83) | 5 (83) | 2 (67) | — |
| SRC | 5 (83) | 1 | — | 4 (80) | — |
| MG | 3 (75) | — | 3 | — | — |
| APS | 1 | 1 | — | — | — |
| mHTN | 1 | — | 1 | — | — |
| TTP | 1 | — | — | 1 | — |
| Drug | 1 | 1 | — | — | — |
Values are n (%) unless specified. P values denote overall differences across the 3 clusters. Significant P values indicate that at least 1 cluster differs from the others.
ADAMTS13, a disintegrin and metalloproteinase with a thrombospondin type 1 motif, member 13; aHUS, atypical hemolytic uremic syndrome; APS, antiphospholipid syndrome; BMT, bone marrow transplant; CFB, complement factor B; CFH, complement factor H; CFHR, complement factor H-related protein; CFI, complement factor I; MG, monoclonal gammopathy; mHTN, malignant hypertension; MPD, myeloproliferative disorder; SLE, systemic lupus erythematosus; SRC, scleroderma renal crisis; THBD, thrombomodulin; TMA, thrombotic microangiopathy; TNF, tumor necrosis factor; TTP, thrombotic thrombocytopenic purpura; VEGF, vascular endothelial growth factor; VUS, variant of uncertain significance.
Regarding the distinction between primary aHUS and mHTN, etiologic assignment was based on clinical predominance: cases were classified as mHTN when severe hypertension was the dominant and presenting feature without prior evidence or strong suspicion of complement dysregulation. However, we acknowledge that, in the absence of routine complement testing, some patients with undiagnosed aHUS may have been included in the mHTN group—a possibility supported by recent literature indicating high rates of complement gene variants in similar populations.
Gemcitabine (n=11; including 1 with abraxane and olaratumab, 1 with abraxane, and 1 with carboplatin and paclitaxel); VEGF inhibitors (n=6; including 3 bevacizumab, 1 with pazopanib and cabozantinib, and 1 with sorafenib); calcineurin inhibitors (n=3); 6-mercaptopurine (n=1); chemotherapy (n=1); and TNF-α inhibitor (n=1).
VEGF inhibitors (n=7 bevacizumab; including 1 with doxorubicin, 1 with niraparib, 1 with doxorubicin and atezolizumab); gemcitabine (n=6; including 1 with paclitaxel); chemotherapy (n=4); calcineurin inhibitors (n=2); tyrosine kinase inhibitors (n=1; dasatinib and nolitinib); aromasin and afinitor (n=1).
Gemcitabine (n=2; including 1 with cisplatin); tyrosine kinase inhibitors (n=2; including 1 with imatinib and 1 with palbociclib); calcineurin inhibitors (n=2); VEGF inhibitor (n=1 bevacizumab); cabazitaxel (n=1); cocaine (n=1); and TNF-α inhibitor (n=1).
In cluster 1, genetic testing was available for 11 patients. Of them, 8 had identifiable variants potentially related to TMA pathogenesis. Among 5 patients diagnosed with aHUS, 2 had pathogenic variants in CFH, 1 had a homozygous deletion of CFHR1, 1 carried a CD46 risk haplotype, and 1 had a combination of a pathogenic variant in CFI along with CFHR3/CFHR1 deletions. One patient with APS was found to carry a VUS in C3. Additionally, 1 patient with drug-induced TMA carried a pathogenic CFH variant, and 1 with SRC had a VUS in CFH combined with a CD46 haplotype.
In cluster 2, genetic testing data were available for 12 patients, of whom 9 had variants potentially contributing to TMA pathogenesis. Among 5 patients diagnosed with aHUS, 2 carried the CD46 risk haplotype, 1 carried the CFH-H3 haplotype, 1 had a VUS in CFI, and another had a VUS in CFB. Three patients with MG had genetic abnormalities including a pathogenic CFH variant, a pathogenic CFI variant, and CFHR3/CFHR1 deletions, respectively. Additionally, 1 patient with mHTN–associated TMA carried the CFH-H3 risk haplotype.
In cluster 3, genetic testing was available for 11 patients, of whom 7 had variants potentially associated with TMA pathogenesis. Among 2 patients with aHUS, 1 carried a VUS in CFH, and the other had a VUS in THBD. Among 4 patients with SRC, 2 had VUS in ADAMTS13 in combination with CFHR3/CFHR1 deletions, 1 had a VUS in C3, and 1 carried a VUS in a CFHR5. Additionally, 1 patient with TTP was found to have a VUS in CFHR5.
Consensus Clustering
We identified 3 distinct clusters using the consensus clustering approach (Supplemental Figure 1, available online at https://www.mcpdigitalhealth.org/). The consensus matrix heatmap showed optimal separation with 3 clusters, as indicated by the highest mean consensus scores: 78 for cluster 1, 92 for cluster 2, and 113 for cluster 3 (Supplemental Figure 2, available online at https://www.mcpdigitalhealth.org/). The cumulative distribution function curve also supported the 3-cluster solutions, showing a plateau in the middle segment-indicative of high clustering stability (Supplemental Figure 3, available online at https://www.mcpdigitalhealth.org/). These convergent results across multiple evaluation metrics consistently support the 3-cluster model as the most stable and most effective for distinguishing patient subgroups.
Clinical Features Across Clusters
Key phenotypic distinctions are summarized in standardized differences plots (Supplemental Figure 4, available online at https://www.mcpdigitalhealth.org/) and detailed in Table 1. Cluster 1 was female predominant (n=50, 64%), compared with 54% (n=50) in cluster 2 and 41% (n=46) in cluster 3. Patients in cluster 2 represented the oldest subgroup (54±18 years) and exhibited the highest eGFR (25 mL/min/1.73 m2; IQR, 14-42). In contrast, cluster 3 comprised the youngest patients (45±14 years), was male predominant (n=67, 59%), and had the highest prevalence of prior chronic kidney disease (n=65, 58%). This cluster also showed the most severe clinical profile, including the highest systolic and diastolic blood pressures (202±44 and 126±32 mm Hg, respectively), the lowest hemoglobin (9.3±1.4 g/dL), the highest frequency of schistocytes (n=81, 72%), and elevated erythrocyte sedimentation rate (42±11 mm/h), and the lowest eGFR (10 mL/min/1.73 m2; IQR, 7-16).
Distribution of Etiologies Across Clusters
Thrombotic microangiopathy etiologies differed significantly across clusters (Table 2). Cluster 1 was dominated by drug-induced TMA (29%, n=23), nearly half of which were gemcitabine-related (11/23; 48%), followed by lupus-associated TMA (12%, n=9) and post–bone marrow transplant (BMT)–associated TMA (10%, n=8) cases.
Cluster 2 also featured a predominance of drug-induced TMA (23%, n=21), with one-third (7/21; 33%) linked to the vascular endothelial growth factor (VEGF) inhibitor bevacizumab. Monoclonal gammopathy–associated (11%, n=10) and post-BMT (9%, n=8) TMAs were frequent, whereas pregnancy-related and myeloproliferative disorder–associated TMAs occurred almost exclusively in this cluster, although their absolute frequencies were low (n=7 and 5; 8% and 5%, respectively).
In contrast, malignant hypertension overwhelmingly dominated in cluster 3, accounting for 66% (n=75) of cases. Drug-induced TMA was rare (9%, n=10), and post-BMT, pregnancy-related, or myeloproliferative disorder–associated TMA were virtually absent. The prevalence of SRC, aHUS, and antiphospholipid syndrome–associated TMA was comparable across clusters.
Genetic Testing Findings
Among 34 patients with genetic testing data, 24 (71%) had mutations potentially related to TMA pathogenesis (Table 2). The overall rate of genetic findings was 80% (12/15) in aHUS, with 83% (5/6) in both clusters 1 and 2, and 67% (2/3) in cluster 3. Notably, 83% (5/6) of patients with SRC had genetic abnormalities. Additionally, the underlying mutation profiles varied across the 3 clusters; cluster 1 was characterized by a predominance of CFH mutations, whereas clusters 2 and 3 displayed more heterogeneous mixes (Supplemental Table 2, available online at https://www.mcpdigitalhealth.org/).
Histologic Features Across Clusters
Renal biopsies were comprehensively evaluated by light microscopy, immunofluorescence, and electron microscopy, revealing distinct histologic patterns across clusters (Table 3, Figure 2, Supplemental Figure 5-7, available online at https://www.mcpdigitalhealth.org/). Cluster 1 exhibited broad immunoactivity, with high positivity for IgM, κ, λ, and C3 (83%-97%), as well as IgA, fibrinogen, C1q, IgG, and albumin staining (50%-70%). In contrast, clusters 2 and 3 showed attenuated immunofluorescence activity: IgM and C3 were positive in 40%-60% of cases, κ or λ light chain positive staining in less than 10%, and IgA virtually absent (3%-4%). IgG and C1q positivity was limited (10%-19%).
Table 3.
Patients’ Kidney Pathology Characteristics
| Variables | Total (N=283) | Cluster 1 (n=78) | Cluster 2 (n=92) | Cluster 3 (n=113) | P |
|---|---|---|---|---|---|
| Light microscopy | |||||
| Segmental glomerulosclerosis | 107 (38) | 32 (41) | 29 (32) | 45 (40) | .31 |
| Wrinkling of capillary tufts | 115 (41) | 23 (29) | 18 (20) | 74 (65) | .0001 |
| Mesangiolysis | 130 (46) | 37 (47) | 66 (72) | 27 (24) | .0001 |
| Double contours | 157 (56) | 57 (73) | 76 (83) | 24 (21) | .0001 |
| Foam cells | 30 (11) | 5 (6) | 15 (16) | 10 (9) | .09 |
| Fragmented red blood cells | 87 (31) | 20 (26) | 24 (26) | 43 (38) | .09 |
| Fibrin thrombi in glomerular capillaries | 85 (30) | 28 (36) | 32 (35) | 25 (22) | .06 |
| Platelet thrombi | 7 (3) | 3 (4) | 2 (2) | 2 (2) | .67 |
| Marginated neutrophils | 30 (11) | 11 (14) | 12 (13) | 7 (6) | .13 |
| Focal karyorrhectic debris | 10 (4) | 5 (6) | 1 (1) | 4 (4) | .15 |
| Reactive endothelium | 14 (5) | 1 (1) | 2 (2) | 11 (10) | .009 |
| Endothelial cell enlargement | 58 (21) | 21 (27) | 26 (28) | 11 (10) | .001 |
| Subendothelial lucency | 42 (15) | 11 (14) | 18 (20) | 13 (12) | .27 |
| Hilar arteriolar thrombus | 28 (10) | 5 (6) | 8 (9) | 15 (13) | .26 |
| Fibrin thrombi | 114 (40) | 28 (36) | 19 (21) | 67 (59) | .0001 |
| Immune thrombi | 1 (0.4) | 1 (1) | 0 (0) | 0 (0) | — |
| Mucoid intimal edema | 139 (49) | 34 (44) | 20 (22) | 85 (75) | .0001 |
| Fibrinoid necrosis of vessel wall | 1 (0.4) | 0 (0) | 1 (1) | 0 (0) | — |
| Intimal hyperplasia with inflammatory cells | 5 (2) | 0 (0) | 2 (2) | 3 (3) | — |
| Onion-skin hyperplastic changes | 74 (26) | 8 (10) | 11 (12) | 55 (49) | .0001 |
| Arteriolosclerosis | 34 (12) | 6 (8) | 14 (15) | 14 (12) | .30 |
| Arteriosclerosis | .0001 | ||||
| None | 195 (69) | 58 (74) | 48 (52) | 89 (79) | |
| Mild | 20 (7) | 6 (8) | 11 (12) | 3 (3) | |
| Moderate | 31 (11) | 7 (9) | 20 (22) | 4 (4) | |
| Severe | 37 (13) | 7 (9) | 13 (14) | 17 (15) | |
| Nodular hyalinosis | 14 (5) | 3 (4) | 5 (5) | 6 (5) | .86 |
| Interstitial edema | 40 (14) | 10 (13) | 15 (16) | 15 (13) | .77 |
| Tubular injury | .0001 | ||||
| None | 17 (6) | 7 (9) | 8 (9) | 2 (2) | |
| Mild | 192 (68) | 62 (79) | 68 (74) | 62 (55) | |
| Moderate | 29 (10) | 3 (4) | 8 (9) | 18 (16) | |
| Severe | 45 (16) | 6 (8) | 8 (9) | 31 (27) | |
| Red blood cell casts | 18 (6) | 6 (8) | 7 (8) | 5 (4) | .54 |
| Immunofluorescence staining | .0001 | ||||
| IgM | 153 (54) | 72 (92) | 35 (38) | 46 (41) | |
| IgG | 77 (27) | 46 (59) | 10 (11) | 21 (19) | |
| IgA | 43 (15) | 39 (50) | 3 (3) | 4 (4) | |
| κ light chain | 90 (32) | 75 (96) | 4 (4) | 11 (10) | |
| λ light chain | 96 (34) | 76 (97) | 8 (9) | 12 (11) | |
| C3 | 164 (58) | 65 (83) | 42 (46) | 57 (50) | |
| C1q | 68 (24) | 45 (58) | 11 (12) | 12 (11) | |
| Albumin | 117 (41) | 54 (69) | 22 (24) | 41 (36) | |
| Fibrinogen | 107 (38) | 44 (56) | 23 (25) | 40 (35) | |
| Electronic microscopy | |||||
| Endothelialitis | 1 (0.4) | 0 (0) | 1 (1) | 0 (0) | — |
| Subendothelial nongranular, flocculent deposits | 21 (7) | 16 (21) | 4 (4) | 1 (0.9) | .0001 |
| Endothelial injury | 149 (53) | 41 (53) | 65 (71) | 43 (38) | .0001 |
| Subendothelial lucency | 193 (68) | 57 (73) | 74 (80) | 62 (55) | .0003 |
| Mesangiolysis | 71 (25) | 26 (33) | 31 (34) | 14 (12) | .0002 |
| Foot process effacement | .003 | ||||
| None | 12 (4) | 1 (1) | 5 (5) | 6 (5) | |
| Mild | 110 (39) | 25 (32) | 30 (33) | 55 (49) | |
| Moderate | 74 (26) | 22 (28) | 20 (22) | 32 (28) | |
| Severe | 87 (31) | 30 (38) | 37 (40) | 20 (18) | |
| Fibrin tactoids | 31 (11) | 10 (13) | 15 (16) | 6 (5) | .03 |
| Double contours | 113 (40) | 38 (49) | 65 (71) | 10 (9) | .0001 |
| Platelet thrombi | 11 (4) | 1 (1) | 7 (8) | 3 (3) | — |
Values are n (%). P values denote overall differences across the 3 clusters. Significant P values indicate that at least 1 cluster differs from the others.
Figure 2.
Distinct kidney pathologies across the 3 clusters. Heatmap displaying the percentage distribution of individual histologic features across each cluster. Color gradient ranges from 0% (green) to 100% (red), with 50% represented by black, indicating the midpoint of prevalence. Each cell reflects the proportion of cases within a cluster exhibiting the specified histologic parameter.
On light microscopy, cluster 1 exhibited a moderate prevalence of glomerular and vascular injury: capillary tuft wrinkling (n=23, 29%), mesangiolysis (n=37, 47%), fibrin thrombi (n=28, 36%), and mucoid intimal edema (n=34, 44%). Onion-skin arteriolar hyperplasia (n=8, 10%), moderate arteriosclerosis (n=7, 9%), and severe tubular injury (n=6, 8%) were infrequent. Electron microscopy revealed notable endothelial injury, including subendothelial nongranular flocculent deposits in 21% (n=16)—the highest among clusters—alongside endothelial cell swelling (n=41, 53%) and double contour formation (n=38, 49%). These features, together with the extensive immunofluorescence positivity, suggest an immune-associated microangiopathy with chronic endothelial remodeling.
Cluster 2 was characterized by the most prominent mesangiolysis (n=66, 72%) of all clusters. Other light microscopic findings, including capillary tuft wrinkling (n=18, 20%), fibrin thrombi (n=19, 21%), and mucoid intimal edema (n=20, 22%), were less frequent than in cluster 3. Onion-skin changes and moderate arteriosclerosis were observed in 11 (12%) and 20 (22%) of cases, respectively. Electron microscopy highlighted severe chronic endothelial injury: endothelial cell damage and double contours in 65 (71%) each, but subendothelial deposits in only 4 (4%). This profile suggests chronic toxic or drug-induced endothelial stress, rather than immune-complex deposition.
Cluster 3 exhibited the most acute and destructive vascular phenotype. Capillary tuft wrinkling (n=74, 65%), fibrin thrombi (n=67, 59%), mucoid intimal edema (n=85, 75%), and onion-skin hyperplastic arteriolar changes (n=55, 49%) were significantly more prevalent than in other clusters, with severe tubular injury in 27% (n=31) of cases. Electron microscopy revealed minimal immune activity—subendothelial deposits in only 1 (0.9%) and low rates of double contours (n=10, 9%) and endothelial injury (n=43, 38%)—suggesting a severe vascular lesion process.
Exploratory Outcome Analysis Findings
Clinical outcome analysis was restricted in the 116 internal patients owing to data availability (Supplemental Table 3, available online at https://www.mcpdigitalhealth.org/). Overall, 52 (45%) developed renal failure. The incidence was significantly higher in cluster 3 (80%, 24/30) than that in cluster 1 (31%, 12/39) and cluster 2 (34%, 16/47; P<.0001). Mortality was observed in 43 (37%), with no statistically significant difference across clusters (P=.08).
The composite outcome occurred in 59% (23/39) in cluster 1, 51% (24/47) in cluster 2, and 90% (27/30) in cluster 3 (P<.001). Kaplan–Meier analysis also found significant differences in both renal failure–free and composite event–free survival across clusters (log-rank test P<.0001 for both) (Supplemental Figure 8, available online at https://www.mcpdigitalhealth.org/). Using cluster 1 as the reference, cluster 3 had a markedly higher risk of the composite outcome (hazard ratio, 2.5; 95% CI, 1.4-4.5; P=.001).
Treatments received across the 3 clusters are presented in Supplemental Table 4 (available online at https://www.mcpdigitalhealth.org/). Overall, 67 (58%) of patients received immunosuppression or eculizumab, with no significant differences among clusters, reflecting real-world practice where these therapies are used broadly in TMA. Drug withdrawal was most common in clusters 1 and 2 (11/39, 28% and 9/47, 19%) than that in cluster 3 (7%, 2/30), showing a trend toward significance (P=.06). This aligns with our clustering results that drug-induced TMA was enriched in clusters 1 and 2. Antihypertensive therapy was more often used in cluster 3 (53%, 16/30), significantly more than in clusters 1 and 2 (8/39, 21% and 8/47, 17%; P=.002), consistent with the high proportion of malignant hypertension in this cluster.
Comparison of baseline characteristics between patients with and without available outcome data revealed significant differences in several variables such as eGFR, blood pressure, proteinuria, and hematuria (Supplemental Table 5, available online at https://www.mcpdigitalhealth.org/), indicating that the outcome findings are not representative of the entire study population. These findings should be considered exploratory and hypothesis generating; the observed differences in outcomes across clusters require validation in a prospectively collected, fully followed cohort before prognostic inferences can be made.
Discussion
We performed unsupervised clustering analysis using routinely available clinicopathologic parameters at presentation. Importantly, we excluded specialized tests with prolonged turnaround times because our objective was to develop a framework that could be applied during initial clinical evaluation. To our knowledge, this is the first study to systematically incorporate renal histopathology—traditionally viewed as nonspecific in the context of TMA13—as a central variable for disease subtyping. The analysis identified 3 reproducible clusters with distinct clinical and renal histologic features (Supplemental Table 6, available online at https://www.mcpdigitalhealth.org/) and outcomes. Although there was not perfect concordance between clusters and specific etiologies, our findings suggest that each cluster may reflect an immune-associated, toxic-related, or severe vascular lesion–related process.
Cluster 1 comprised mostly females of intermediate age, with relatively preserved blood pressure and moderate disease severity. The dominant etiologies in this group—drug-induced TMA (notably gemcitabine), lupus, and post-BMT—share a common underpinning of immune activation or dysregulation.14, 15, 16 Correspondingly, immunofluorescence revealed widespread positivity for immune molecules. Electron microscopy found subendothelial deposits and double contours, reflecting chronic glomerular endothelial remodeling. These findings collectively suggest a phenotype likely driven by systemic immune-associated injury, consistent with the pathophysiology of lupus nephritis, posttransplant immune activation, and certain drug hypersensitivity reactions. Importantly, this group exhibited the strongest histologic and serologic evidence of immune involvement, potentially pointing toward early use of immunosuppressive therapy.
Cluster 2 had relatively preserved renal function, with the highest platelet counts, suggesting comparatively less acute clinical severity at presentation. It was also dominated by drug-induced TMA, especially related to anti-VEGF agents such as bevacizumab, as well as calcineurin inhibitors and tyrosine kinase inhibitors. These drugs are known to disrupt endothelial integrity through direct cytotoxic, antiangiogenic, or metabolic pathways.17,18 Importantly, VEGF signaling is known to regulate local complement inhibitory proteins, such as factor H, and disruption of this axis can predispose to complement-mediated endothelial injury.19 Similarly, calcineurin inhibitors impair factor H binding and cofactor activity, potentially unmasking complement dysfunction. This cluster also included rare etiologies such as monoclonal gammopathy–associated, pregnancy-associated, and myeloproliferative disorders. All have been implicated in endothelial injury via cytokine release or direct toxicity.20,21 Thus, even so-called secondary TMAs may involve dysregulated complement activity, blurring the line between toxic- and immune-associated TMA.22 The biopsy findings suggest a toxic, non–immune-driven pathogenesis—with the highest frequencies of mesangiolysis and endothelial swelling, prominent double contours, minimal subendothelial deposits, and modest positive immunofluorescence staining—indicative of structural damage rather than immune deposition.23
Cluster 3 was distinct in its demographic and clinical presentation: younger age, male predominance, significantly elevated blood pressure, and lowest eGFR. Malignant hypertension was the dominant etiology, accounting for two-thirds of cases, with associated etiologies including aHUS and SRC. This group had the worst outcomes; 80% developed renal failure. Histologically, these patients exhibited classic vascular lesions of pressure-induced injury, severe tubular injury, and minimal subendothelial deposits and low immune activity on immunofluorescence. These findings suggest that the dominant insult is not an immune mechanism but, rather, a severe vascular lesion process, likely driven by extreme intraluminal pressure and shear stress damaging the microvascular endothelium.24 This aligns with prior literature describing TMA in malignant hypertension as a result of red blood cell fragmentation under high shear stress.25, 26, 27 Similar principles apply in TTP (where ADAMTS13 deficiency permits unregulated von Willebrand factor–platelet interactions) and in SRC (where renin-driven vasoconstriction leads to endothelial damage).27, 28, 29, 30 These features across the conditions suggest that cluster 3 may represent a convergence of different upstream mechanisms sharing a common vascular injury pattern and poor prognosis.
Outcome data were available for 16 patients with malignant hypertension (3 in cluster 1, 4 in cluster 2, and 9 in cluster 3). Although numbers are too small for formal statistical comparison, a trend consistent with cluster-based risk stratification was observed: the composite outcome occurred in 1 of 3 (33%) of patients in cluster 1, 2 of 4 (50%) in cluster 2, and 8 of 9 (89%) in cluster 3. This suggests that, even within the same broad etiology, phenotypic clustering may identify subgroups with meaningfully different outcomes—supporting the notion that the cluster provides prognostic information beyond phenotyping alone.
Genetic variants play a central role in TMA pathogenesis, especially in complement-mediated forms such as aHUS. Prior studies have shown that approximately 50%-60% of patients with complement-mediated TMA carry pathogenic or susceptibility variants, which influence both disease expression and risk of recurrence.31 In our cohort, among the 34 patients who underwent testing, 71% had abnormalities potentially contributing to TMA. The yield of genetic testing was high in SRC and aHUS, with 83% and 80% tested having a mutation identified, respectively. Notably, most of malignant hypertension-attributed cases did not undergo genetic testing. However, prior studies suggest a high rate of complement abnormalities in this population, with 92% having primary aHUS.32,33 Therefore, the frequent overlap between malignant hypertension and complement-mediated injury—particularly aHUS—introduces ambiguity. These data highlight the diagnostic uncertainty in malignant hypertension-associated TMA, especially in the absence of genetic workup. Without comprehensive complement testing, we cannot definitively exclude aHUS or other underlying complement-mediated etiologies in cluster 3.
The identification of 3 distinct TMA phenotypes holds direct implications for clinical management, offering a potential framework for early therapeutic considerations at presentation. Patients in cluster 1—characterized by immune-associated features—would be considered for early immunosuppressive strategies. In contrast, cluster 2 should prompt urgent review and discontinuation of potential offending agents; in cases with suspected complement dysregulation, such as those with complement gene variants, early complement testing and possible complement blockade may be warranted. Cluster 3, dominated by malignant hypertension and severe vascular lesions with minimal immune staining, prioritizes aggressive antihypertensive therapy and volume control over immunosuppression, although complement testing remains relevant given potential overlapping presentations such as aHUS or SRC. Although traditional etiologic categories were distributed across all 3 clusters, this overlap underscores that TMA is a histologic end point with multiple convergent pathways.34 Our clustering model does not merely recapitulate existing etiologic classifications but integrates routinely available clinicopathologic data to offer an early pathogenic insight. By translating phenotypic patterns into actionable diagnostic and therapeutic priorities, this approach can help bridge the gap between biopsy results and targeted intervention, even in the absence of definitive genetic or serologic results. The clustering approach would be particularly valuable when a patient presents with multiple potential TMA triggers (Supplemental Table 7, available online at https://www.mcpdigitalhealth.org/). Thus, the clustering advances understanding by adding a phenotypic layer to traditional etiologic classification. Notably, treatment patterns in the internal cohort showed that clinical management, although not guided by our clustering framework, partly aligned with the phenotypic distinctions we identified. However, the substantial overlap in immunosuppressive and complement-blocking therapies across clusters reflects the complexity of TMA management in practice and underscores the need for more precise tools that integrate clinical, genetic, and kidney biopsy data to guide therapeutic selection.
This is one of the largest biopsy-confirmed TMA cohorts analyzed to date, although some subgroups remain small, limiting statistical power to detect subtle distinctions. A key limitation of this study is the incomplete and clinically driven application of genetic testing, which was performed primarily in cases with high suspicion of complement-mediated disease. This introduces substantial ascertainment bias, limiting our ability to accurately determine etiology or estimate the true prevalence of genetic variants across the 3 clusters. Consequently, robust genotype–phenotype correlations and definitive validation of the proposed mechanistic clusters were not feasible; mechanistic inferences remain speculative. In light of this constraint, the phenotypic labels—immune-associated, toxic-related, and severe vascular lesion–associated processes—should be viewed as clinicopathologic patterns reflecting predominant modes of injury rather than as molecularly defined disease entities.
Conclusion
Using unsupervised machine learning on integrated clinical and histologic data, we identified 3 clinicopathologic phenotypes that may suggest, but not prove, underlying immune-, toxic-, or severe vascular lesion–related drivers. These clinicopathologic clusters may help inform early diagnostic prioritization and therapeutic consideration, supporting the role of renal biopsy in providing pathogenic insight before advanced testing results are available. However, definitive pathogenic assignment requires systematic molecular and genetic profiling to guide individualized management. Future prospective studies are needed to map genetic variants onto phenotypic groups, refine their pathogenic underpinnings, and evaluate whether phenotype-guided therapy improves outcomes in well-defined patient populations. Additionally, this work is hypothesis generating, and the current approach is not yet suitable for routine practice; a clinically usable classifier with reduced variables requires future development.
Potential Competing Interests
None.
Ethics Statement
This retrospective study was approved by Mayo Clinic Institutional Review Board (21-007513).
Acknowledgments
Drs Simon Moubarak and Jing Miao contributed equally to this work.
Footnotes
Supplemental material can be found online at https://www.mcpdigitalhealth.org/. Supplemental material attached to journal articles has not been edited, and the authors take responsibility for the accuracy of all data.
Supplemental Online Material
References
- 1.George J.N., Nester C.M. Syndromes of thrombotic microangiopathy. N Engl J Med. 2014;371(7):654–666. doi: 10.1056/NEJMra1312353. [DOI] [PubMed] [Google Scholar]
- 2.Moake J.L. Thrombotic microangiopathies. N Engl J Med. 2002;347(8):589–600. doi: 10.1056/NEJMra020528. [DOI] [PubMed] [Google Scholar]
- 3.Brocklebank V., Wood K.M., Kavanagh D. Thrombotic microangiopathy and the kidney. Clin J Am Soc Nephrol. 2018;13(2):300–317. doi: 10.2215/CJN.00620117. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Lusco M.A., Fogo A.B., Najafian B., Alpers C.E. AJKD atlas of renal pathology: thrombotic microangiopathy. Am J Kidney Dis. 2016;68(6):e33–e34. doi: 10.1053/j.ajkd.2016.10.006. [DOI] [PubMed] [Google Scholar]
- 5.Benz K., Amann K. Thrombotic microangiopathy: new insights. Curr Opin Nephrol Hypertens. 2010;19(3):242–247. doi: 10.1097/MNH.0b013e3283378f25. [DOI] [PubMed] [Google Scholar]
- 6.Jokiranta T.S. HUS and atypical HUS. Blood. 2017;129(21):2847–2856. doi: 10.1182/blood-2016-11-709865. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Palma L.M.P., Sridharan M., Sethi S. Complement in secondary thrombotic microangiopathy. Kidney Int Rep. 2021;6(1):11–23. doi: 10.1016/j.ekir.2020.10.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Bommer M., Wolfle-Guter M., Bohl S., Kuchenbauer F. The differential diagnosis and treatment of thrombotic microangiopathies. Dtsch Arztebl Int. 2018;115(19):327–334. doi: 10.3238/arztebl.2018.0327. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Shatzel J.J., Taylor J.A. Syndromes of thrombotic microangiopathy. Med Clin North Am. 2017;101(7):395–415. doi: 10.1056/NEJMra1312353. [DOI] [PubMed] [Google Scholar]
- 10.Zheng Z., Waikar S.S., Schmidt I.M., et al. Subtyping CKD patients by consensus clustering: the Chronic Renal Insufficiency Cohort (CRIC) Study. J Am Soc Nephrol. 2021;32(3):639–653. doi: 10.1681/ASN.2020030239. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Monti S., Tamayo P., Mesirov J., Golub T. Consensus clustering: a resampling-based method for class discovery and visualization of gene expression microarray data. Mach Learn. 2003;52:91–118. doi: 10.1023/A:1023949509487. [DOI] [Google Scholar]
- 12.Mclnnes L., Healy J., Melville J. UMAP: uniform manifold approximation and projection for dimension reduction. Preprint. Posted online February 9, 2020 doi: 10.48550/arXiv.1802.03426. arXiv 1802.03426. [DOI] [Google Scholar]
- 13.Gallan A.J., Chang A. A new paradigm for renal thrombotic microangiopathy. Semin Diagn Pathol. 2020;37(3):121–126. doi: 10.1053/j.semdp.2020.01.002. [DOI] [PubMed] [Google Scholar]
- 14.Blasco M., Guillen E., Quintana L.F., et al. Thrombotic microangiopathies assessment: mind the complement. Clin Kidney J. 2021;14(4):1055–1066. doi: 10.1093/ckj/sfaa195. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Martis N., Jamme M., Bagnis-Isnard C., et al. Systemic autoimmune disorders associated with thrombotic microangiopathy: a cross-sectional analysis from the French National TMA registry: systemic autoimmune disease-associated TMA. Eur J Intern Med. 2021;93:78–86. doi: 10.1016/j.ejim.2021.05.040. [DOI] [PubMed] [Google Scholar]
- 16.Turner J.L., Reardon J., Bekaii-Saab T., Cataland S.R., Arango M.J. Gemcitabine-associated thrombotic microangiopathy: response to complement inhibition and reinitiation of gemcitabine. Clin Colorectal Cancer. 2017;16(2):e119–e122. doi: 10.1016/j.clcc.2016.09.004. [DOI] [PubMed] [Google Scholar]
- 17.Goldberg R.J., Nakagawa T., Johnson R.J., Thurman J.M. The role of endothelial cell injury in thrombotic microangiopathy. Am J Kidney Dis. 2010;56(6):1168–1174. doi: 10.1053/j.ajkd.2010.06.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Jhaveri K.D., Flombaum C.D., Kroog G., Glezerman I.G. Nephrotoxicities associated with the use of tyrosine kinase inhibitors: a single-center experience and review of the literature. Nephron Clin Pract. 2011;117(4):c312–c319. doi: 10.1159/000319885. [DOI] [PubMed] [Google Scholar]
- 19.Keir L.S., Firth R., Aponik L., et al. VEGF regulates local inhibitory complement proteins in the eye and kidney. J Clin Invest. 2017;127(1):199–214. doi: 10.1172/JCI86418. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Buttner-Herold M., Sticht C., Wiech T., Porubsky S. Renal disease associated with myeloproliferative neoplasms and myelodysplastic syndrome/myeloproliferative neoplasms. Histopathology. 2021;78(5):738–748. doi: 10.1111/his.14282. [DOI] [PubMed] [Google Scholar]
- 21.Ravindran A., Go R.S., Fervenza F.C., Sethi S. Thrombotic microangiopathy associated with monoclonal gammopathy. Kidney Int. 2017;91(3):691–698. doi: 10.1016/j.kint.2016.09.045. [DOI] [PubMed] [Google Scholar]
- 22.Teoh C.W., Riedl Khursigara M., Ortiz-Sandoval C.G., et al. The loss of glycocalyx integrity impairs complement factor H binding and contributes to cyclosporine-induced endothelial cell injury. Front Med (Lausanne) 2023;10 doi: 10.3389/fmed.2023.891513. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Ren Y., Wang H., Qin H., et al. Vascular endothelial growth factor expression in peripheral blood of patients with pregnancy induced hypertension syndrome and its clinical significance. Pak J Med Sci. 2014;30(3):634–637. doi: 10.12669/pjms.303.4558. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Gallo G., Volpe M., Savoia C. Endothelial dysfunction in hypertension: current concepts and clinical implications. Front Med (Lausanne) 2021;8 doi: 10.3389/fmed.2021.798958. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Li W., Lian R., Li Y., et al. Kidney outcomes of malignant hypertension-associated thrombotic microangiopathy in patients with and without IgA nephropathy: a propensity score-matched analysis. Clin Kidney J. 2025;18(3) doi: 10.1093/ckj/sfaf017. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Mitaka H., Yamada Y., Hamada O., Kosaka S., Fujiwara N., Miyakawa Y. Malignant hypertension with thrombotic microangiopathy. Intern Med. 2016;55(16):2277–2280. doi: 10.2169/internalmedicine.55.6332. [DOI] [PubMed] [Google Scholar]
- 27.Shibagaki Y., Fujita T. Thrombotic microangiopathy in malignant hypertension and hemolytic uremic syndrome (HUS)/thrombotic thrombocytopenic purpura (TTP): can we differentiate one from the other? Hypertens Res. 2005;28(1):89–95. doi: 10.1291/hypres.28.89. [DOI] [PubMed] [Google Scholar]
- 28.Basyal B., Finnigan N.A. Scleroderma and Renal Crisis. StatPearls. 2025 [PubMed] [Google Scholar]
- 29.Chrabaszcz M., Malyszko J., Sikora M., et al. Renal involvement in systemic sclerosis: an update. Kidney Blood Press Res. 2020;45(4):532–548. doi: 10.1159/000507886. [DOI] [PubMed] [Google Scholar]
- 30.Zander C.B., Cao W., Zheng X.L. ADAMTS13 and von Willebrand factor interactions. Curr Opin Hematol. 2015;22(5):452–459. doi: 10.1097/MOH.0000000000000169. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Agarwal S., Sukumar S., Masias C., Cataland S. Genetic landscape of thrombotic microangiopathies with focus on plasminogen mutation variants. Blood. 2023;142(Suppl 1):4002. doi: 10.1182/blood-2023-187248. [DOI] [Google Scholar]
- 32.Cavero T., Aunon P., Caravaca-Fontan F., et al. Thrombotic microangiopathy in patients with malignant hypertension. Nephrol Dial Transplant. 2023;38(5):1217–1226. doi: 10.1093/ndt/gfac248. [DOI] [PubMed] [Google Scholar]
- 33.Cavero T., Arjona E., Soto K., et al. Severe and malignant hypertension are common in primary atypical hemolytic uremic syndrome. Kidney Int. 2019;96(4):995–1004. doi: 10.1016/j.kint.2019.05.014. [DOI] [PubMed] [Google Scholar]
- 34.Hanna R.M., Henriksen K., Kalantar-Zadeh K., Ferrey A., Burwick R., Jhaveri K.D. Thrombotic microangiopathy syndromes-common ground and distinct frontiers. Adv Chronic Kidney Dis. 2022;29(2):149–160.e1. doi: 10.1053/j.ackd.2021.11.006. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.