Intraglomerular Dysfunction Predicts Kidney Failure in Type 2 Diabetes

Glomerular filtration rate (GFR), considered the best overall index of kidney function, is a criterion in defining and staging of chronic kidney disease (1–4). The key physiological determinants of GFR are renal plasma flow (RPF) and measures of intraglomerular hemodynamic function including intraglomerular pressure (P_GLO) and the balance in vascular resistance between the afferent (R_A) and efferent (R_E) arterioles (R_A-to-R_E ratio). Nephroprotection conferred by renin-angiotensin-aldosterone system (RAAS) and sodium–glucose cotransporter 2 (SGLT2) blockers has been attributed, at least in part, to effects on these hemodynamic parameters (5–8). Indeed, the substantial nephroprotection demonstrated with SGLT2 inhibitors in cardiovascular and kidney outcome trials of people with type 2 diabetes (T2D) has been linked to the drug class’ influence on tubuloglomerular feedback and intraglomerular hemodynamic alterations (9–11).

P_GLO and R_A-to-R_E ratio, not directly measurable in humans, can be estimated with equations developed by Gomez et al. (12,13) Accurate estimation requires gold standard measures of GFR and RPF. Whether changes in intraglomerular hemodynamic function based on these equations explain end-stage kidney disease (ESKD) risk remains unknown because of the absence of longitudinal studies with repeated gold standard measures of GFR and RPF and sufficient duration to capture their effects on ESKD.

In this study, we examined the role of intraglomerular hemodynamic function in the progression to ESKD in an American Indian cohort with T2D. Also examined was the relationship between intraglomerular hemodynamic function and diabetic kidney disease (DKD) structural injury from research kidney biopsies. We hypothesized that elevated P_GLO and R_A-to-R_E ratio associate with greater risk of ESKD. Our findings provide a framework for understanding the role of intraglomerular hemodynamic abnormalities in the natural history of DKD progression, and we suggest how certain drugs may mitigate DKD.

Among Pima Indian individuals from the Gila River Indian Community in Arizona there is a high prevalence of T2D and a high incidence of ESKD due to T2D. Between 1965 and 2007, Pima Indian individuals participated in a longitudinal study of the natural history of diabetes and its complications. In 1988, informative subsets of individuals from this population were selected to undergo more detailed longitudinal studies of kidney hemodynamic parameters in DKD (14–18), and many of these individuals continued to be followed.

The current study was conducted in a subset of 237 adult participants who had repeated measures of GFR and RPF. The baseline exam for each participant was their first exam with a measurement of GFR and RPF. Of these 237 participants, 169 were enrolled in a clinical trial of losartan effects on DKD progression (by design, during the trial 84 were randomly assigned to losartan and 85 to placebo). Among them, 32 had initiated their follow-up before enrollment in this trial and 111 were invited to undergo a research kidney biopsy at the end of this trial (14). Of the 111 participants included in the research kidney biopsy study, 61 were randomized to losartan and 50 to placebo.

ESKD onset, defined as the initiation of kidney replacement therapy or death from DKD if the participant opted out of dialysis, was ascertained independently in all study participants through 31 December 2019. The study was approved by the Institutional Review Board of the National Institute of Diabetes and Digestive and Kidney Diseases. Each participant provided written informed consent.

Blood pressure was measured twice while the participant was resting in the supine (2.7%) or seated position (97.3%), and the measurements were averaged. Mean arterial pressure (MAP) was calculated as follows: (2 × diastolic blood pressure + systolic blood pressure) / 3. GFR and RPF were measured annually throughout follow-up by the urinary clearance of iothalamate and p-aminohippurate (PAH), respectively. High-performance liquid chromatography was used to measure the concentrations of iothalamate and PAH. Filtration fraction (FF) was calculated as GFR / RPF. Urine albumin concentration was measured by nephelometric immunoassay and urine creatinine by a modified Jaffé reaction until 1 August 2011, when this assay was replaced by the enzymatic method (19). Albumin excretion was assessed by the urine albumin–to–creatinine ratio (ACR). Urine albumin concentration below the detection limit of the assay (≤6.8 mg/L) was set to 6.8 mg/L in the analyses.

Measurement techniques have previously been described (14). Briefly, unbiased random sampling of kidney biopsy tissue sections provided digital images for quantitative morphometric estimates of DKD structural parameters by observers masked to the clinical data. Predefined structural parameters quantified included glomerular basement membrane (GBM) width, cortical interstitial fractional volume, glomerular mesangial fractional volume, glomerular volume, glomerular filtration surface density, podocyte number per glomerulus, podocyte foot process width, percent podocyte detachment, percent fenestrated endothelium, and total filtration surface per glomerulus. The equation to calculate the percentage of globally sclerotic glomeruli accounted for the difference in size and, consequently, the probability of encountering sclerotic versus nonsclerotic glomeruli in random cross sections (15).

Indirect intraglomerular hemodynamic parameters, including P_GLO and R_A-to-R_E ratio, were estimated with use of equations derived by Gomez et al., which are based on animal experimental data (Supplementary Fig. 1) (12,13). The following parameters were calculated:

where RBF is renal blood flow and RVR is renal vascular resistance.

Assumptions imposed by Gomez equations are as follows: 1) intrarenal vascular resistances are divided into afferent, postglomerular, and efferent; 2) hydrostatic pressures within the renal tubules, venules, Bowman’s space, and interstitium (P_Bow) are in equilibrium of 10 mmHg; 3) glomerulus is in filtration disequilibrium; and 4) the gross filtration coefficient (K_F) is 0.0867 mL/s/mmHg given a normal kidney. The Gomez equations were also used to calculate a second set of intraglomerular hemodynamic parameters assuming K_F = 0.1012 mL/s/mmHg for patients with diabetes. MAP (mmHg), RPF (mL/s), GFR (mL/s), and total protein (g/dL) were used to calculate R_E (dyne * s/cm⁵) and R_A (dyne * s/cm⁵), P_GLO (mmHg), filtration pressure across glomerular capillaries (ΔPF) (mmHg), and glomerular oncotic pressure (π_G) (mmHg). Calculations are as follows: ΔPF = GFR/ultafiltration coefficient (K_FG). The πG from the plasma protein mean concentration (CM) within the capillaries: CM = total protein (TP)/FF × ln(1/1 – FF). π_G = 5 × (CM – 2). P_GLO = ΔPF + P_Bow + π_G. R_A and R_E were estimated using principles of Ohm’s law, where 1,328 is the conversion factor to dyne * s/cm⁵: R_A = [(MAP − P_GLO)/RBF] × 1,328, R_E = [GFR/K_FG × (RBF − GFR)] × 1,328, where RBF is renal blood flow.

Patient characteristics are expressed as mean ± SD, with median (25th–75th percentile) for skewed distributions; qualitative variables are presented as frequencies and percentages. Positively skewed parameters were log transformed as appropriate.

Cox proportional hazards regression was used to assess effects of parameters on the ESKD risk. A time-varying model was performed to account for variation of intraglomerular parameters over time: adjustment was made for baseline age, sex, and diabetes duration and for the time-dependent variables GFR, HbA_1c, BMI, and RAAS inhibitor use. The hazard ratio (HR) and 95% CI of ESKD was expressed per 1 SD. Each model was tested for log-linearity and proportionality assumptions. Because sex did not fulfill the proportional hazards assumption, sex stratification was used. Since other antihypertensive agents and nonsteroidal anti-inflammatory drugs (NSAIDs) may influence intraglomerular hemodynamics, we conducted two sensitivity analyses by considering time-varying NSAIDs and antihypertensive exposure in the model.

Heterogeneity in the hemodynamic parameters’ trajectories was investigated using latent class mixed modeling with the R lcmm package (20). We estimated the latent class linear mixed models that separate the population into homogeneous subgroups of individuals according to their trajectory. The linear mixed model expressed the observed hemodynamic parameters of each participant in each latent class at the time of measurement. Patient characteristics in the different latent classes were compared with assessment of the clinical relevance of this classification using traditional statistical tests (ANOVA or Kruskal-Wallis test, χ² test).

Time to event was plotted as Kaplan-Meier ESKD-free survival curves according to latent classes and compared using the log-rank test. Since glycemia is an important DKD risk factor and latent classes groups had significant differences in diabetes duration, Kaplan-Meier curves were also plotted from diabetes diagnosis instead of first visit. Cross-sectional relationships of hemodynamic parameters from the visit closest to the kidney biopsy were assessed by Spearman correlations and graphically illustrated by partial residual regression plots. Partial correlation analysis was used to study these relationships after adjustment for effects of age, sex, diabetes duration, BMI, HbA_1c, and MAP. ACR is highly correlated with the underlying structural lesions. Because we considered ACR as an intermediate phenotype between renal hemodynamics and renal outcomes, it was not included in these models for adjustment.

Statistical analyses were performed with SAS, version 9.4 (SAS Institute, Cary, NC); R statistical software, version 3.6.3; and GraphPad Prism software, version 8.0. P values <0.05 were considered statistically significant. Analyses were considered exploratory and hypothesis generating, and we did not adjust for multiple comparisons.

The data sets generated and analyzed during this study are available from the corresponding author upon reasonable request.

Sixty-nine percent of participants were women. Mean ± SD baseline age was 42 ± 11 years and diabetes duration 11 ± 7 years. Mean GFR was 153 ± 45 mL/min, and median ACR was 36 mg/g (interquartile range 13–139). During a median follow-up of 17.5 years (10.6–20.5), 2,226 kidney clearances of iothalamate and PAH were performed (median 10 per participant [range 1–21]). ESKD occurred in 69 (29%) participants. Participant characteristics stratified by ESKD during follow-up are summarized in Table 1. At baseline, 8% of the participants were receiving an RAAS inhibitor. During follow-up, 171 of 237 participants received 1,063 person-years of RAAS blocker treatment during 2,348 person-years of observation.

Compared with participants without ESKD, those developing ESKD had significantly longer diabetes duration, higher albuminuria and HbA_1c, and lower BMI and total serum protein at baseline. Participants developing ESKD also had significantly lower FF and higher RPF (Table 1). After multivariable adjustment for baseline age, sex, diabetes duration, and BMI, HbA_1c, RAAS inhibitor treatment, and GFR in time-varying models, increases in RPF, RBF, and R_A-to-R_E ratio and decreases in FF and P_GLO predicted ESKD risk (Fig. 1). Sensitivity analyses with NSAIDs agents or antihypertensive drugs considered as time-dependent covariate did not modify our findings (data not shown).

Latent class analyses identified two distinct temporal patterns for P_GLO and R_A-to-R_E ratio (class 1 and class 2 [Fig. 2 and Supplementary Table 1]). Whereas class 2 trajectories for P_GLO and R_A-to-R_E ratio were stable over time, class 1 trajectories were not and were therefore termed as “unstable” for descriptive purposes. P_GLO in class 1 was significantly higher at baseline than in class 2 (mean ± SD 69.1 ± 5.8 vs. 62.8 ± 8.9 mmHg, P < 0.0001) and then declined progressively, and the R_A-to-R_E ratio in class 1 increased over time. The unstable trajectories in class 1 participants were associated with considerably higher risk of incident ESKD than the stable trajectories in class 2 (HR for P_GLO 2.96, 95% CI 1.45–6.02, P = 0.003; HR for R_A-to-R_E ratio 4.60, 95% CI 2.55–8.31, P < 0.0001) (Fig. 3). Participant characteristics by these classes for P_GLO and R_A-to-R_E ratio are shown in Table 2. Among the 46 participants with either high-risk P_GLO or R_A-to-R_E ratio patterns, only a single participant exhibited both high-risk trajectories of intraglomerular hemodynamic dysfunction.

To examine associations between kidney structure and hemodynamic parameters, we used the values from the exam closest to the kidney biopsy (median time between biopsy and measures 2.4 months [range 0.2–32.3]). Partial Spearman rank correlations between hemodynamic parameters and structural parameters from that biopsy, after adjustment for age, sex, HbA_1c, and MAP, are shown in Fig. 4 and Supplementary Table 2. Lower GFR, FF, and P_GLO correlated significantly with greater structural lesions, including lower filtration surface density (surface/glomerular volume), fewer endothelial fenestrations/peripheral endothelial surface area, greater glomerular basement membrane width, and higher mesangial and cortical interstitial fractional volumes. Higher R_A-to-R_E ratio correlated with the same structural lesions. Notably, P_GLO and R_A-to-R_E ratio explained a greater degree of the variance of the structural lesions than GFR and RPF.

This study delineates, for the first time, specific intraglomerular hemodynamic parameters that predict ESKD in persons with T2D using gold standard measures of GFR and RPF. These findings were based on serial measurements of GFR and RPF performed over several decades in this American Indian cohort. Increasing R_A-to-R_E ratio associated strongly with incident ESKD independent of GFR and traditional risk factors. Likewise, higher P_GLO followed by a progressive decline in this parameter conferred greater risk of incident ESKD. Furthermore, each of these intraglomerular hemodynamic parameters correlated strongly with the glomerular structural lesions that best predict loss of GFR and progression to ESKD. Interestingly, participants with progressive DKD typically exhibited one or the other of these abnormal trajectories, but not both, suggesting that each of these trajectories represents a distinct phenotype and may reflect different mechanisms of DKD progression.

The intraglomerular hemodynamic parameters in this study, including P_GLO and R_A-to-R_E ratio, are not directly measurable in humans and must be estimated from the Gomez equations (12,13). Accurate estimates require gold standard measures of GFR and RPF like those performed here. Since most drugs known to mitigate DKD risk affect intraglomerular hemodynamic function, the current study provides a clinical framework for better understanding of mechanisms underlying these hemodynamic effects.

The two drug classes that offer the greatest nephroprotection in DKD are the RAAS and SGLT2 inhibitors. RAAS blockade is thought to confer nephroprotection, in part, by efferent arteriolar vasodilation (5–8). Similarly, treatment with SGLT2 inhibitors promotes postglomerular vasodilation and reduced P_GLO (17,21). Nevertheless, the demonstrated benefits of these drugs cannot be specifically ascribed to their short-term actions on kidney hemodynamics, highlighting the uncertainty about the role of intraglomerular hemodynamic changes on DKD progression. The current study directly addresses this uncertainty by examining longitudinal relationships between intraglomerular hemodynamic function, DKD lesions, and ESKD. Mechanisms by which these drugs mitigate DKD progression, however, may be multifaceted, remain under intensive study, and cannot be directly addressed by the current analysis (22).

The trajectories of P_GLO and R_A-to-R_E ratio in our prospective data emphasize the temporal complexity of these intraglomerular hemodynamic changes and the potential constraints of extrapolating findings from small short-term mechanistic trials to explain hard kidney outcome benefits in large long-term clinical trials. For example, participants from the high-risk unstable trajectory of P_GLO had significantly higher baseline P_GLO than their peers from the low-risk stable trajectory, followed by a more rapid decline of P_GLO. Other studies using Gomez equations or invasive measures of renal arterial pressure and flow velocity established that P_GLO is 45–55 mmHg in healthy control subjects and is, on average, 10 mmHg higher in individuals with T2D (12,23). By these criteria, participants in the high-risk trajectory of P_GLO exhibited early intraglomerular hypertension. Therefore, our findings may support the notion that attenuation of intraglomerular hypertension early in the course of DKD could alter the subsequent P_GLO trajectory and mitigate ESKD risk. However, lowering of normal or suppressed P_GLO later in the course of DKD may have the opposite effect and magnify the risk of ESKD. These changes may also reflect worsening glomerular lesions, such as mesangial expansion, which limit glomerular capillary luminal space (24). The two high-risk P_GLO and R_A-to-R_E ratio phenotypes identified in adults with T2D may be used to enrich clinical trials with individuals at greater risk of DKD progression. Since these phenotypes exhibit distinct hemodynamic patterns, they may respond differently to current and novel therapies. Molecular interrogation of kidney tissues from individuals with these phenotypes may uncover different targetable pathways for drug development, thus facilitating personalized medicine.

This study has important strengths and limitations. Strengths include a large number of adults with T2D with up to 31 years of prospective data collected by a rigorous protocol of comprehensive and regular phenotyping. Serial measurements of GFR and RPF by iothalamate and PAH clearance allowed longitudinal examination of how trajectories of intraglomerular hemodynamic function relate to ESKD. Additionally, research kidney biopsies enabled examination of relationships between intraglomerular hemodynamic parameters and DKD morphometric parameters. To our knowledge, this is the only study with long-term longitudinal data on intraglomerular hemodynamic function, renal structure, and ESKD. However, its uniqueness also precludes external validation of our findings. Nevertheless, risk factors for ESKD identified in the Pima Indian population with T2D have been consistently replicated in other populations, arguing that findings in this population are generalizable (25). Additionally, the cumulative incidence of ESKD in Pima Indians with T2D is similar to that observed in the predominantly Caucasian patients with type 1 diabetes from the Joslin Clinic but far higher than in the predominantly Caucasian patients with T2D from the Mayo Clinic. We attributed this difference to the younger age at onset of diabetes and the lower death rate from coronary heart disease in the Pima Indians than in the Caucasians with T2D (26,27). In the present analysis, there were more women than men who participated, which may have confounded our findings due to potential sexual dimorphism of intraglomerular hemodynamic function. Another limitation is that Gomez equation parameters are derived from whole-kidney GFR and RPF in lieu of direct measurements. Additionally, average estimates of the ultrafiltration coefficient derived from micropuncture studies were used in the analysis. However, our P_GLO values are consistent with those obtained using invasive renal arterial pressure and flow velocity measurements in participants with T2D (23). Furthermore, Gomez equation parameters do not reflect single-nephron function. Ascertainment of changes in single-nephron function would require serial research kidney biopsies and reliable assessments of nephron number, the latter subject to biases and technical challenges (28). Finally, the time between measurement of GFR and RPF and the kidney biopsies ranged between 0.2 to 32.3 months. Structural changes of diabetic kidney injury are usually indolent and can take several years to manifest (typically >3–5 years). For this reason, we contend that potential confounding of the temporal variability of hemodynamic and morphometric parameters is likely modest.

In conclusion, trajectories of R_A-to-R_E ratio and P_GLO, independent of GFR, associated strongly with incident ESKD and the glomerular structural lesions that best predict ESKD progression. Identifying the metabolic and molecular mechanisms underlying these hemodynamic imbalances may provide new therapeutic targets for DKD in T2D.

Acknowledgments. The authors acknowledge the work of Lois I. Jones, Enrique Diaz, Bernadine Waseta, Camille Waseta, Julie Paul, and Joey de Keizer.

Funding. Financial support for this work provided by the American Diabetes Association (Clinical Science Award 1-08-CR-42) and by the Intramural Research Program of the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK). P.B. receives salary and research support from NIDDK (DK116720, DK114886), JDRF (2-SRA-2019-845-S-B, 3-SRA-2017-424-M-B), Boettcher Foundation, Center for Women’s Health Research at University of Colorado, and the Section of Endocrinology Department of Pediatrics, and Barbara Davis Center for Diabetes at University of Colorado School of Medicine.

Duality of Interest. P.B. has acted as a consultant for AstraZeneca, Bayer, Bristol-Myers Squibb, Boehringer Ingelheim, Eli Lilly, Sanofi, Novo Nordisk, and Horizon Pharma. P.B. serves on the advisory boards for AstraZeneca, Bayer, Boehringer Ingelheim, Novo Nordisk, and XORTX. P.J.S. has acted as a consultant for AstraZeneca. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. P.J.S., H.C.L., R.G.N., and P.B. wrote the manuscript and researched data. R.G.N. designed the study. M.M. and B.N. were responsible for morphometric analyses and contributed to discussion and reviewed and edited the manuscript. E.G. and S.R. assisted in analyses and contributed to discussion and reviewed and edited the manuscript. P.J.S., R.G.N., and P.B. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.