About

Harold I. Feldman, MD, MSCE, is an Emeritus Professor of Biostatistics and Epidemiology at the University of Pennsylvania's Perelman School of Medicine. He serves as an Attending Physician at the Hospital of the University of Pennsylvania and is a Senior Fellow at the Leonard Davis Institute of Health Economics. Additionally, he is a Senior Scholar at the Center for Clinical Epidemiology and Biostatistics and the Director of the Center for Clinical Epidemiology and Biostatistics at the University of Pennsylvania. His research expertise includes epidemiology, patient-oriented research, and kidney disease, with clinical expertise in nephrology and internal medicine. Dr. Feldman's work involves significant contributions to understanding chronic kidney disease biomarkers, blood pressure management in kidney disease, and acute kidney injury, among other areas. His scholarly activities include numerous publications in reputable journals, reflecting his focus on advancing clinical epidemiology and health economics.

Research topics

• Medicine
• Internal medicine
• Endocrinology
• Intensive care medicine
• Urology
• Biology
• Gastroenterology
• Linguistics
• Emergency medicine

Selected publications

• Five-Year Risk Prediction Models for Peripheral Artery Disease in Patients With Chronic Kidney Disease

  Kidney Medicine · 2026-04-03

  articleOpen access

  Rationale & Objective: Patients with chronic kidney disease (CKD) have an increased risk of peripheral artery disease (PAD), yet no PAD risk-prediction models currently exist for this population. We developed and internally validated 5-year PAD risk-prediction models for individuals with CKD. Study Design: Prospective cohort study. Setting & Participants: 3,076 patients with CKD without PAD from the Chronic Renal Insufficiency Cohort (CRIC) study. Exposure: Clinically available variables, ankle-brachial index (ABI), and non-routinely measured cardiovascular disease biomarkers assessed at baseline. Outcomes: New-onset adjudicated clinical PAD event or an ABI ≤0.9 at an annual follow-up visit. Analytical Approach: Cox proportional hazards models were applied to estimate 5-year risk of incident PAD from baseline. Model performance was assessed by discrimination, calibration, and net reclassification improvement. All models were internally validated using Monte Carlo cross-validation. Results: <0.001). The biomarker-enhanced model achieved an AUC of 0.724 (95% CI, 0.711-0.745), which was not significantly different from the clinical model with ABI. Both the clinical model with ABI and the biomarker-enhanced model were well calibrated and improved reclassification of non-events compared with the ABI model (19.7%; 95% CI, 17.5-22.0% and 22.2%; 95% CI, 19.8-24.5%, respectively). Limitations: Lack of external validation. Conclusions: A PAD risk prediction model that combines clinical variables with ABI improves the identification of patients with CKD at high risk for PAD better than ABI alone.

  PublisherDOI

• Hemodynamics are associated with subsequent lumen remodeling and clinical maturation of hemodialysis arteriovenous fistula

  Scientific Reports · 2025-02-19 · 7 citations

  articleOpen access

  Abstract The pathogenesis of arteriovenous fistula (AVF) maturation failure is unclear. We evaluated the associations of wall shear stress (WSS) with subsequent AVF remodeling and clinical maturation using regression models in this prospective cohort study. Participants underwent duplex ultrasound at postoperative Day 1, Week 2, and Week 6 to measure AVF blood flow rate and diameter of the draining vein and proximal artery. The median vein WSS of 602 participants decreased from Day 1 to Week 6 (from 33.4 to 21.6 dyne/cm 2 ) but did not change noticeably for the artery (from 58.4 to 55.1 dyne/cm 2 ). WSS was positively associated with subsequent lumen expansion, with doubling of Day-1 WSS presaging a 9% (95% confidence interval (CI) 5%-14%; P &lt; 0.001) greater Day 1-to-Week 6 increase in vein lumen cross-sectional area and a 5% (95% CI: 1%-10%; P = 0.020) greater increase in artery lumen area. The odds of unassisted clinical maturation increased by 45% (95% CI: 11%-89%; P = 0.006) with each doubling of Day-1 vein WSS, and by 82% (95% CI: 39%-250%; P &lt; 0.001) with each doubling of Day-1 artery WSS. These findings show that WSS was positively associated with subsequent lumen expansion and AVF unassisted clinical maturation.

  PublisherOA PDFDOI

• Family History in the Context of CKD

  Journal of the American Society of Nephrology · 2025-03-11 · 3 citations

  article

  Key Points Family history of kidney disease was not associated with kidney disease progression in the context of established CKD. Family history of diabetes was a risk factor of CKD progression independently of diabetes status, polygenic risk, and traditional risk factors. Background A family history of health conditions may reflect shared genetic and/or environmental risk. It is not well known to what extent family history affects outcomes among patients with CKD. In this study, we investigated the associations of family history of CKD, diabetes, and other conditions with common comorbidities and kidney disease progression among patients with CKD. Methods We performed an observational study of two prospective CKD cohorts, 2573 adults and children from the Cure Glomerulopathy Network and 3939 Chronic Renal Insufficiency Cohort adult participants. Self-reported first-degree family history of CKD, diabetes, and other common diseases was tested for associations with the risk of comorbidities and CKD progression using multivariable models. Results Family history of common comorbid conditions was associated with higher risk of these conditions in the context of CKD, including approximately by over three-fold for diabetes (adjusted odds ratio [OR], 3.37; 95% confidence interval [CI], 2.73 to 4.15), 48% for cancer (adjusted OR, 1.48; 95% CI, 1.05 to 2.09), and 69% for cardiovascular disease (adjusted OR, 1.69; 95% CI, 1.36 to 2.10 in combined cohorts). While polygenic risk score (PRS) for CKD was associated with kidney disease progression (adjusted hazards ratio, 1.11; 95% CI, 1.06 to 1.16 in combined cohorts), family history of kidney disease was not an independent risk factor of disease progression in the context of existing CKD. By contrast, family history of diabetes was significantly associated with a higher risk of CKD progression independently of diabetes occurrence or PRS for diabetes (adjusted hazards ratio, 1.19; 95% CI, 1.05 to 1.35 in combined cohorts). Conclusions Broad collection of family history in the context of CKD improved clinical risk stratification. Family history of diabetes was consistently associated with a higher risk of CKD progression independently of diabetes status or PRS for diabetes in both cohorts.

  PublisherDOI

• Deoxycholic Acid and Cognitive Impairment and Decline in the Chronic Renal Insufficiency Cohort (CRIC)

  Kidney Medicine · 2025-04-25

  articleOpen access

  <h3>Rationale & Objective</h3> Cognitive impairment is common in chronic kidney disease (CKD). The secondary bile acid, deoxycholic acid (DCA), is associated with cognitive impairment and Alzheimer's dementia among older adults without CKD. Whether DCA is associated with cognitive impairment and decline in CKD is unknown. <h3>Study Design</h3> Cross-sectional and longitudinal multivariable-adjusted regression analyses. <h3>Setting & Participants</h3> 2,836 CRIC Study participants; 699 CRIC Cognitive (COG) Study participants. <h3>Exposure</h3> Fasting serum DCA levels measured at visit 5 (ie, baseline). <h3>Outcomes</h3> Modified Mini-Mental State Examination (3MS) in the main CRIC cohort and domain-specific cognitive tests in the CRIC COG cohort: Trail Making Test Parts A and B, Category Fluency, Buschke Selective Reminding, and Boston Naming. Cognitive impairment was defined as test score>1 standard deviation worse than the mean test score. <h3>Results</h3> Mean age 59±10 years, 45% female, and 39% Black. In the overall cohort, in cross-sectional analyses, there was no association between DCA and cognitive impairment by 3MS in after adjustment for demographics and clinical factors (prevalence ratio doubling DCA, 1.00; 95% CI, 0.95-1.06; n=2,836). In longitudinal analyses, DCA was associated with decline (mean annual percent change in 3MS per doubling DCA, −0.13; 95% CI, −0.28 to−0.02) but not with incident impairment (n=2,836; follow-up of 8.6±3.9 years). Among CRIC COG Study participants, in cross-sectional analyses, DCA was associated with cognitive impairment based on Category Fluency (prevalence ratio per doubling DCA, 1.14; 95% CI, 1.02-1.27) but not with other specific-domain cognitive tests (n=698-699). In CRIC COG longitudinal analyses, DCA was not associated with decline or incident cognitive impairment (n=538-574). <h3>Limitations</h3> No adjustment for inflammation, no stool DCA, 3MS may lack specificity. <h3>Conclusion</h3> Among individuals with CKD stages 2-4, higher DCA levels were independently associated with prevalent cognitive impairment in Category Fluency. The association between DCA and progressive cognitive impairment assessed by 3MS was small and likely not clinically significant. <h3>Plain-Language Summary</h3> Cognitive impairment is common in chronic kidney disease (CKD). Deoxycholic acid (DCA) is a secondary bile acid that is associated with cognitive impairment in older adults without CKD, but its association with cognitive impairment and decline in patients with CKD is unknown. We examined whether DCA was associated with baseline or later cognitive impairment among 2,836 participants from a cohort of adults with CKD stages 2-4. Although higher DCA levels were associated with one measure of cognitive impairment at baseline, the association with progressive cognitive impairment was small and likely not clinically significant.

  PublisherDOI

• Renin-Angiotensin System Inhibitors in Patients With Nonproteinuric Chronic Kidney Disease and Kidney Outcomes: Findings From the CRIC Study

  Kidney Medicine · 2025-10-25 · 1 citations

  articleOpen access

  <h3>Rationale & Objective</h3> The kidney-protective benefits of renin-angiotensin system inhibitors (RASIs; ie, angiotensin-converting enzyme inhibitors and angiotensin receptor blockers) in nonproteinuric chronic kidney disease (CKD) are unclear. We aimed to evaluate kidney outcomes in adults with nonproteinuric CKD treated with RASIs versus other antihypertensive medications. <h3>Study Design</h3> Using data from the Chronic Renal Insufficiency Cohort study, a prospective cohort study, we evaluated the association of RASI use with kidney outcomes and survival. Secondary analyses included censoring with drug discontinuation and a time-updated RASI approach (ie, cumulative exposure). <h3>Setting & Participants</h3> Individuals with <0.5 g/d of proteinuria (via 24-hour urine collection or spot test) on ≥1 antihypertensive medication. Participants with heart failure were excluded. <h3>Exposure</h3> Patient-reported use of RASIs versus non-RASI antihypertensive medication. <h3>Outcomes</h3> (1) CKD progression (halving of estimated glomerular filtration rate, dialysis, or transplant) and (2) all-cause mortality. <h3>Analytical Approach</h3> Inverse probability of treatment weighting and Cox proportional hazards modeling; marginal structural models. <h3>Results</h3> Among the 2,664 participants, the mean age was 62 years, 46% were female, and the mean estimated glomerular filtration rate was 50 mL/min/1.73 m². There were 457 kidney events (29/1,000 person-years) in RASI users versus 129 (23/1,000 person-years) in the comparator. Treatment with RASIs was not associated with CKD progression in the baseline analysis (adjusted hazard ratio [HR], 1.01; 95% CI, 0.81-1.25) and drug discontinuation analysis (adjusted HR, 0.92; 95% CI, 0.68-1.25). The cumulative exposure approach showed a higher risk of CKD progression for low RASI users (<50% of follow-up) versus nonusers but no higher risk of CKD progression among high RASI users (≥50% of follow-up) versus nonusers. RASI users had a lower mortality risk in the drug discontinuation analysis (adjusted HR, 0.64; 95% CI, 0.50-0.82) and a nonsignificantly lower risk among participants receiving treatment with RASIs ≥75% of follow-up versus nonusers. <h3>Limitations</h3> Residual confounding cannot be ruled out. Medication use was patient-reported, increasing the potential for misclassification. <h3>Conclusions</h3> For people with nonproteinuric CKD, RASIs may not be kidney-protective but may still confer a mortality benefit for hypertension management. <h3>Plain-Language Summary</h3> Chronic kidney disease is a leading cause of illness and death. Although the majority of individuals with chronic kidney disease do not have proteinuria, clinical trials have historically excluded nonproteinuric chronic kidney disease. Currently, no treatments are known to slow kidney disease progression in this population. Our study examined whether renin-angiotensin system inhibitors, a class of antihypertensive medication known to be kidney-protective in proteinuric kidney diseases, provides similar benefits in nonproteinuric chronic kidney disease. In a long-term observational study, we used multiple methods to account for changing medication use over time. Although we found no evidence that renin-angiotensin system inhibitors improve kidney outcomes compared with other blood pressure medications, we observed some evidence of a mortality benefit.

  PublisherOA PDFDOI

• Glycolytic lactate in diabetic kidney disease

  JCI Insight · 2024-06-09 · 37 citations

  articleOpen access

  Lactate elevation is a well-characterized biomarker of mitochondrial dysfunction, but its role in diabetic kidney disease (DKD) is not well defined. Urine lactate was measured in patients with type 2 diabetes (T2D) in 3 cohorts (HUNT3, SMART2D, CRIC). Urine and plasma lactate were measured during euglycemic and hyperglycemic clamps in participants with type 1 diabetes (T1D). Patients in the HUNT3 cohort with DKD had elevated urine lactate levels compared with age- and sex-matched controls. In patients in the SMART2D and CRIC cohorts, the third tertile of urine lactate/creatinine was associated with more rapid estimated glomerular filtration rate decline, relative to first tertile. Patients with T1D demonstrated a strong association between glucose and lactate in both plasma and urine. Glucose-stimulated lactate likely derives in part from proximal tubular cells, since lactate production was attenuated with sodium-glucose cotransporter-2 (SGLT2) inhibition in kidney sections and in SGLT2-deficient mice. Several glycolytic genes were elevated in human diabetic proximal tubules. Lactate levels above 2.5 mM potently inhibited mitochondrial oxidative phosphorylation in human proximal tubule (HK2) cells. We conclude that increased lactate production under diabetic conditions can contribute to mitochondrial dysfunction and become a feed-forward component to DKD pathogenesis.

  PublisherOA PDFDOI

• Matrix Metalloproteinase-2 and CKD Progression: The Chronic Renal Insufficiency Cohort (CRIC) Study

  Kidney Medicine · 2024-06-07 · 7 citations

  articleOpen access

  Rationale & Objective: Matrix metalloproteinase 2 (MMP-2) plays an important role in the development of fibrosis, the final common pathway of chronic kidney disease (CKD). This study aimed to assess the relationship between repeated measures of MMP-2 and CKD progression in a large, diverse prospective cohort. Study Design: In a prospective cohort of Chronic Renal Insufficiency Cohort (CRIC) participants (N = 3,827), MMP-2 was measured at baseline. In a case-cohort design, MMP-2 was additionally measured at year 2 in a randomly selected subcohort and cases of estimated glomerular filtration rate (eGFR) halving or kidney replacement therapy (KRT) (N = 1,439). Setting & Participants: CRIC is a multicenter prospective cohort of adults with CKD. Exposure: MMP-2 measured in plasma at baseline and at year 2. Outcomes: A composite kidney endpoint (KRT/eGFR halving). Analytical Approach: Weighted Cox proportional hazards models for case-cohort participants. Results: = 0.09) after adjusting for covariates. The relationship of persistently elevated MMP-2 was modified by levels of inflammation, with a 2.6 times higher rate of the composite kidney endpoint in those with high-sensitivity C-reactive protein < 2.5 g/dL at study entry. Heterogeneity of effect was found with proteinuria, with a baseline MMP-2 level of ≥300 ng/mL associated with an increased risk of the composite kidney endpoint (HR, 1.30; 95% CI, 1.09-1.54) only with proteinuria ≥ 442 mg/g. Limitations: The observational study design limits causal interpretation. Conclusions: Elevated MMP-2 is associated with CKD progression, particularly among those with low inflammation and those with proteinuria. Future investigations are warranted to confirm the reduction in risk of CKD progression among these subgroups of patients with CKD.

  PublisherOA PDFDOI

• Abstract 4135307: Integrating genomics and proteomics to identify causal proteins and biologic pathways for incident atrial fibrillation in chronic kidney disease

  Circulation · 2024-11-12

  article

  Introduction/ Background: Individuals with chronic kidney disease (CKD) are at an increased risk for developing atrial fibrillation (AF). Research Objective: By integrating genomics and proteomics, we sought to identify proteins and biologic pathways causally linked to the development of AF in the CKD population. Methods / Approach: We measured 4638 unique plasma proteins using SomaScan v.4.0 in participants (ppts) with CKD and no prior AF from 2 cohorts: Chronic Renal Insufficiency Cohort (CRIC, n=2654) and Atherosclerosis Risk in Communities Cohort (ARIC, n=1326). Associations of individual proteins with incident AF were analyzed using Cox regression after adjustment for demographics, comorbidities, and kidney function. Among proteins independently associated with incident AF in both cohorts, we tested for causality using two-sample Mendelian Randomization (MR). Previously published genome wide association studies (GWAS) were used to identify both protein quantitative loci (cis-pQTL) and summary statistics for incident AF. We then used Ingenuity Pathway Analysis to compare proteins and pathways associated with both incident AF and an increase in left atrial (LA) size across serial echocardiograms in CRIC. Results / Data: Over the 5-year analytic period, incident AF occurred in 150 ppts from CRIC and 140 ppts from ARIC. Eight proteins were independently associated with AF in both cohorts: NT-proBNP, SVEP1, TAGLN, TFF3, WAP4, MMP-12, CILP2 and NELL-1. Of these, 3 were significant in MR analysis (Figure 1): CILP2 and MMP-12 maintain structural integrity of the extracellular matrix in the heart and modulate the fibrotic response; NELL-1 is responsible for cell signaling in kidney disease and cardiovascular development. In addition, leading biologic pathways implicated in both development of AF and increase in LA size included systemic fibrosis (p=1.7x10 -4 for AF and 2.3 x10 -4 for LA); regulation of insulin-like growth factor (p=2.3 x10 -4 for AF and p=4.3 x10 -7 for LA) and extracellular matrix degradation (p=5.5 x10 -4 for AF and p=4.2 x10 -4 for LA). Conclusions: Trans-omics analyses identified 3 proteins causally associated with developing AF in the CKD population. Also, several pathways involved with cardiac remodeling and fibrosis are common to the development of AF and LA remodeling.

  PublisherDOI

• Depressive Symptoms, Antidepressants, and Clinical Outcomes in Chronic Kidney Disease: Findings from the CRIC Study

  Kidney Medicine · 2024-02-09 · 12 citations

  articleOpen access

  Rationale & Objective: The extent to which depression affects the progression of chronic kidney disease (CKD) and leads to adverse clinical outcomes remains inadequately understood. We examined the association of depressive symptoms (DS) and antidepressant medication use on clinical outcomes in 4,839 adults with nondialysis CKD. Study Design: Observational cohort study. Setting and Participants: Adults with mild to moderate CKD who participated in the multicenter Chronic Renal Insufficiency Cohort Study (CRIC). Exposure: The Beck Depression Inventory (BDI) was used to quantify DS. Antidepressant use was identified from medication bottles and prescription lists. Individual effects of DS and antidepressants were examined along with categorization as follows: (1) BDI <11 and no antidepressant use, (2) BDI <11 with antidepressant use, (3) BDI ≥11 and no antidepressant use, and (4) BDI ≥11 with antidepressant use. Outcomes: CKD progression, incident cardiovascular disease composite, all-cause hospitalizations, and mortality. Analytic Approach: Cox regression models were fitted for outcomes of CKD progression, incident cardiovascular disease, and all-cause mortality, whereas hospitalizations used Poisson regression. Results: At baseline, 27.3% of participants had elevated DS, and 19.7% used antidepressants. Elevated DS at baseline were associated with significantly greater risk for an incident cardiovascular disease event, hospitalization, and all-cause mortality, but not CKD progression, adjusted for antidepressants. Antidepressant use was associated with higher risk for all-cause mortality and hospitalizations, after adjusting for DS. Compared to participants without elevated DS and not using antidepressants, the remaining groups (BDI <11 with antidepressants; BDI ≥11 and no antidepressants; BDI ≥11 with antidepressants) showed higher risks of hospitalization and all-cause mortality. Limitations: Inability to infer causality among depressive symptoms, antidepressants, and outcomes. Additionally, the absence of nonpharmacological data, and required exploration of generalizability and alternative analytical approaches. Conclusions: Elevated DS increased adverse outcome risk in nondialysis CKD, unattenuated by antidepressants. Additionally, investigation into the utilization and counterproductivity of antidepressants in this population is warranted.

  PublisherOA PDFDOI

• Proteomic Assessment of the Risk of Secondary Cardiovascular Events among Individuals with CKD

  Journal of the American Society of Nephrology · 2024-09-26 · 8 citations

  articleOpen access

  Key Points Machine learning and large-scale proteomics led to a 16-protein secondary cardiovascular risk model in patients with CKD. Hepatic fibrosis and liver X receptor activation represented biologic pathways that link kidney disease and risk of secondary cardiovascular events. An understanding of the circulating proteins associated with secondary cardiovascular events may help to identify novel therapeutic targets. Background Cardiovascular risk models have been developed primarily for incident events. Well-performing models are lacking to predict secondary cardiovascular events among people with a history of coronary heart disease, stroke, or heart failure who also have CKD. We sought to develop a proteomic risk score for cardiovascular events in individuals with CKD and a history of cardiovascular disease. Methods We measured 4638 plasma proteins among 1067 participants from the Chronic Renal Insufficiency Cohort (CRIC) and 536 individuals from the Atherosclerosis Risk in Communities (ARIC) Cohort. All had non–dialysis-dependent CKD and coronary heart disease, heart failure, or stroke at study baseline. A proteomic risk model for secondary cardiovascular events was derived by elastic net regression in CRIC, validated in ARIC, and compared with clinical models. Biologic mechanisms of secondary events were characterized through proteomic pathway analysis. Results A 16-protein risk model was superior to the Framingham Risk Score for secondary events, including a modified score that included eGFR. In CRIC, the annualized area under the receiver operating characteristic curve (area under the curve) within 1–5 years ranged between 0.77 and 0.80 for the protein model and 0.57 and 0.72 for the clinical models. These findings were replicated in the ARIC validation cohort. Biologic pathway analysis identified pathways and proteins for cardiac remodeling and fibrosis, vascular disease, and thrombosis. Conclusions The proteomic risk model for secondary cardiovascular events outperformed clinical models on the basis of traditional risk factors and eGFR.

  PublisherOA PDFDOI

Recent grants

• NIH Grant R29DK042907

  NIH · $559k · 1995

• NIH Grant U01DK060990

  NIH · $50.1M · 2018

• Chronic Kidney Disease (CKD) Biomarkers Consortium Data Coordinating Center

  NIH · $5.5M · 2014–2024

• NIH Grant R01DK044994

  NIH · $1.4M · 2000

• Clinical Research Training in Kidney Disease

  NIH · $3.9M · 2000–2021

Frequent coauthors

• Raymond R. Townsend

  653 shared

• John W. Kusek

  University of Pennsylvania

  634 shared

• Amanda H. Anderson

  Tulane University

  524 shared

• Jiang He

  Central South University

  455 shared

• Radhakrishna R. Kallem

  University of Pennsylvania

  417 shared

• Wei Yang

  Imperial College London

  403 shared

• Alan S. Go

  Stanford University

  383 shared

• Nisha Bansal

  352 shared

Labs

• Biostatistics and EpidemiologyPI

Education

• MSCE

  University of Pennsylvania

  1991