Structural Predictors of Loss of Renal Function in American Indians with Type 2 Diabetes

Abstract

Background and objectives

Diabetes is the leading cause of kidney failure in the United States, but early structural determinants of renal function loss in type 2 diabetes are poorly defined. We examined the association between morphometrically determined renal structural variables and loss of renal function in 111 American Indians with type 2 diabetes who volunteered for a research kidney biopsy at the end of a 6-year clinical trial designed to test the renoprotective efficacy of losartan versus placebo. Participants were subsequently followed in an observational study, in which annual measurements of GFR (iothalamate) initiated during the clinical trial were continued.

Design, setting, participants, & measurements

Renal function loss was defined as ≥40% loss of GFR from the research examination performed at the time of kidney biopsy. Associations with renal function loss were evaluated by Cox proportional hazards regression. Hazard ratios (HRs) were reported per 1-SD increment for each morphometric variable.

Results

Of 111 participants (82% women; baseline mean [±SD] age, 46 years old [±10]; diabetes duration, 16 years [±6]; hemoglobin A1c =9.4% [±2.2]; GFR=147 ml/min [±56]; median albumin-to-creatinine ratio, 41 mg/g [interquartile range, 13–158]), 51 (46%) developed renal function loss during a median follow-up of 6.6 years (interquartile range, 3.1–9.0). Fourteen had baseline GFR <90 ml/min, and three had baseline GFR <60 ml/min. Higher mesangial fractional volume (HR, 2.27; 95% confidence interval [95% CI], 1.58 to 3.26), percentage of global glomerular sclerosis (HR, 1.63; 95% CI, 1.21 to 2.21), nonpodocyte cell number per glomerulus (HR, 1.50; 95% CI, 1.10 to 2.05), glomerular basement membrane width (HR, 1.48; 95% CI, 1.05 to 2.08), mean glomerular volume (HR, 1.42; 95% CI, 1.02 to 1.96), and podocyte foot process width (HR, 1.28; 95% CI, 1.03 to 1.60); lower glomerular filtration surface density (HR, 0.62; 95% CI, 0.41 to 0.94); and fewer endothelial fenestrations (HR, 0.68; 95% CI, 0.48 to 0.95) were each associated with GFR decline after adjustment for baseline age, sex, duration of diabetes, hemoglobin A1c, GFR, and treatment assignment during the clinical trial.

Conclusions

Quantitative measures of glomerular structure predict loss of renal function in type 2 diabetes.

Introduction

Renal function loss (RFL) is an important indicator of progressive diabetic kidney disease and often precedes the appearance of elevated urinary albumin excretion in both type 1 and type 2 diabetes (1–3). Structural determinants of early RFL are well characterized in type 1 diabetes (4,5) but not in type 2 diabetes owing to the heterogeneous nature of the renal lesions in many populations (6–9). Whereas some individuals with type 2 diabetes have patterns characteristic of diabetes injury on kidney biopsy, others do not, and the rate of RFL may differ according to the histologic pattern (7,9,10). Uncertainty regarding the structural determinants of kidney disease in type 2 diabetes is exacerbated by the use of kidney biopsies obtained for clinical reasons (11) and the failure to apply quantitative estimates of renal structure and careful measurements of GFR.

Diabetic kidney disease in Pima Indians is well characterized. The lesions of diabetic glomerulosclerosis parallel those seen in other populations (12), hyperfiltration is an early feature of the disease, and the incidence of ESRD is similar to that reported in type 1 diabetes (13–16). In this study, we analyzed protocol kidney biopsies from 111 Pima Indians with type 2 diabetes to examine the relationship between structural lesions and RFL, defined as ≥40% reduction in GFR from the baseline value. Study participants were followed annually after the kidney biopsy for measurement of GFR by the urinary clearance of iothalamate. A ≥40% decline in GFR was selected because of a recent meta-analysis of 1.7 million participants, including some with diabetes, that found that even declines in eGFR of <57%, which is equivalent to a doubling of serum creatinine concentration, were strongly and consistently associated with the risk of ESRD and death (17). Declines of 30%–40% have been proposed as outcomes for clinical trials in CKD (18), and we selected the stricter ≥40% decline for this study, because the evidence was stronger for this end point and because many participants had normal or elevated GFR at baseline. Key structural-functional relationships among 37 (33%) members of this cohort were described previously (19), and therefore, this study will focus on the structural predictors of RFL.

Materials and Methods

Pima Indians from the Gila River Indian Community participated in a longitudinal study from 1965 to 2007. Diabetes was diagnosed by a 2-hour postload plasma glucose concentration ≥200 mg/dl at biennial research examinations or when the diagnosis was otherwise documented in the medical record. One hundred sixty-nine adult Pima Indians from this longitudinal study who had type 2 diabetes participated in a 6-year clinical trial to evaluate the renoprotective efficacy of losartan (ClinicalTrials.gov no. NCT00340678) (20). The prespecified primary end point was a decline in GFR to ≤60 ml/min or one half the baseline value in participants who entered the study with GFR<120 ml/min. GFR was measured annually throughout the trial. At the end of the trial, a kidney biopsy was performed in 111 of the participants to determine if the treatment was also associated with preservation of normal kidney structure. The participants who underwent kidney biopsies were included in this study, and baseline clinical data for this study were obtained from the renal function test closest to the biopsy. The median time interval between the renal function test and the kidney biopsy was 1.6 months (interquartile range [IQR], 0.9–4.2 months). Thirty-eight (34%) of these tests were performed before the end of randomized treatment; 19 of these participants were assigned to the study drug, and 19 were assigned to placebo. After the clinical trial was completed, all participants were invited to participate in this long–term observational study. Treatment provided during the observational study was prescribed by the participants’ primary care physicians. GFR was measured a median of nine times (range =1–13) after enrollment in this study; three participants had only one additional measurement during follow-up.

This study was approved by the Institutional Review Board of the National Institute of Diabetes and Digestive and Kidney Diseases. Each subject gave written informed consent.

Laboratory Measurements

Urinary albumin was measured by nephelometric immunoassay, and urinary creatinine was measured by a modified Jaffé reaction (21,22). Urinary albumin excretion (albumin-to-creatinine ratio [ACR]) was considered normal if <30 mg/g, microalbuminuria if ≥30 mg/g but <300 mg/g, and macroalbuminuria if ≥300 mg/g. HPLC was used to measure nonradioactive iothalamate for GFR determination (23).

GFR Measurement by Iothalamate Clearance

GFR was measured after an overnight fast (23). At the beginning of each clearance study, the bladder was emptied, a urine sample was collected for measurement of ACR, and a diuresis was initiated by an oral water load. A loading dose of 30% iothalamate (300 mg plus 3 mg/kg for each 1 kg >100 kg) was given intravenously followed by a continuous infusion to maintain the serum concentration constant at about 1.5 mg/dl. After a 60-minute equilibration period, the bladder was again emptied, and four urine collections were made at 20-minute intervals. Blood was drawn to bracket each urine collection, and the concentration of iothalamate was measured in each of these blood and urine samples (23).

Morphometric Methods

Unbiased random sampling of tissue sections provided digital images for study using quantitative morphometric methods to estimate renal structural parameters (20,24). Measurements were performed by an investigator (E.J.W.) who was masked to the clinical data. Parameters included mean glomerular volume, glomerular basement membrane (GBM) width, cortical interstitial fractional volume, mesangial fractional volume per glomerulus, glomerular filtration surface density, total filtration surface per glomerulus, number of nonpodocyte cells (endothelial cells and mesangial cells) per glomerulus, number of podocytes per glomerulus, podocyte foot process width, percentage of podocyte detachment, and percentage of normally fenestrated endothelium (12,19,20,25,26). An equation that accounts for the difference in probability of encountering a sclerotic or nonsclerotic glomerulus in a random cross-section was used to calculate the percentage of sclerotic glomeruli (27). On average, 15±6 glomeruli were examined in each participant by light microscopy and 3±1 were examined in each participant by electron microscopy for the morphometric measurements. Ten grids were placed on light microscopic images of tissue sections for measurement of cortical interstitial fractional volume by point counting. Morphometric variables for each individual were calculated as the mean of all measurements for that individual.

Statistical Analyses

Clinical features of study participants are presented using numbers and percentages for categorical variables, means and SDs for normally distributed variables, and medians and IQRs for skewed data. Differences in morphometric structure by albuminuria level and GFR were examined by the Wilcoxon two–sample test. Associations between clinical characteristics and the morphometric variables were explored by Spearman correlation.

Follow-up began at the renal function test closest to the kidney biopsy and ended in participants who developed RFL at the first visit in which GFR declined by ≥40% relative to baseline and in those who did not develop RFL at the last measurement of kidney function. Individual GFR slopes were also computed as rates of change by simple linear regression and expressed as the annual change in milliliters per minute. We used absolute measurements of GFR in these analyses, because the study included overweight and obese participants, and indexing for body surface area may significantly underestimate their actual GFR (28). Moreover, a change in weight during follow-up would change the body surface area–adjusted GFR value, even in the absence of an actual change in GFR.

Cox proportional hazards regression was used to examine the association between baseline structure and RFL during follow-up. The hazard ratio (HR) was expressed for a 1-SD increment in the distribution of each morphometric variable. Two models are described: (1) univariate and (2) adjusted for baseline age, sex, diabetes duration, hemoglobin A1c, GFR, and treatment assignment during the clinical trial. A glomerulopathy index was developed by standardizing the sum of the standardized values of each morphometric variable significantly associated with RFL after multivariable adjustment. HRs were also computed to examine the association between baseline structure and a decline in GFR slope of ≥5 ml/min per year. Because BP and albuminuria are intermediate variables between renal lesions and RFL, additional adjustment for these variables was not performed. Generalized c statistics were calculated for the glomerulopathy index model, accounting for variable follow-up times (29). Comparison between a nested model that included or excluded this variable was assessed by a likelihood ratio test (30,31). The magnitude of improvement in discrimination of the 6-year risk for RFL with the addition of the index was assessed by relative integrated discrimination improvement (32,33). The 6-year risk was selected, because it approximated the median follow-up time for the study outcome.

Six sensitivity analyses were performed. In the first analysis, GFR slope computed for the period from the initiation of the clinical trial to the research examination performed at the time of kidney biopsy replaced GFR measured at the time of biopsy in the regression analysis. In the second analysis, GFR measured at each research examination at which the participant reported current treatment with a renin-angiotensin system (RAS) inhibitor was adjusted upward by 3.75% to account for the acute effects of initiating treatment with RAS inhibitors (20). These adjusted GFR values were used to define RFL in this analysis. In the third analysis, the RFL outcome was modified to require ≥40% reduction in GFR and a final GFR of <90 ml/min. Seventeen progressors had GFRs≥90 ml/min after experiencing a ≥40% decline and were included as nonprogressors in this analysis. In two other analyses, RFL was defined by either a ≥30% reduction or a ≥57% reduction in GFR. In the last analysis, GFR slope was analyzed as a continuous variable using linear regression.

Results

Clinical characteristics at baseline are shown in Table 1. Fourteen participants had GFR<90 ml/min at baseline; three of them had GFR<60 ml/min. Hyperfiltration, defined by a GFR≥154 ml/min, a value 2 SDs above the mean GFR for Pima Indians with normal glucose tolerance, was present at baseline in 41 participants. Fifty-one participants (46%; 15 normoalbuminuria, 22 microalbuminuria, and 14 macroalbuminuria) developed RFL during a median follow-up of 6.6 years (IQR, 3.1–9.0 years). Twenty-eight (55%) of the participants with RFL had hyperfiltration at baseline, and the mean GFR at baseline was higher in those with RFL than in those without (169 versus 128 ml/min; P<0.001); 52 participants had normal ACR at baseline, 36 had microalbuminuria, and 23 had macroalbuminuria.

Table 1.

Baseline characteristics of 111 Pima Indians (29 men and 82 women) with type 2 diabetes

Characteristic	Mean±SD
Age, yr	46±10
Diabetes duration, yr	16±6
Body mass index, kg/m2	36±8
BP, mmHg
Systolic	124±15
Diastolic	78±9
Hemoglobin A1c, %	9.4±2.2
RAS agent, % taking RAS	65 (59)a
Treatment assignment during clinical trial (placebo, losartan)	50, 61b
GFR, ml/min	147±56
GFR<90 ml/min, GFR=90–153 ml/min, GFR≥154 ml/min	14, 56, 41b
Albumin-to-creatinine ratio, mg/g	41 (13–158)c
Normoalbuminuria, microalbuminuria, macroalbuminuria	52, 36, 23b

The number of patients actually taking renin-angiotensin system (RAS) agents at the time of the kidney biopsy is shown as the treatment assignment of each participant during the clinical trial that immediately preceded the kidney biopsy.

^aNumber (%).

^bNumber.

^cMedian (interquartile range).

Renal structural parameters at baseline are shown in Tables 2 and 3. Participants with elevated albuminuria had greater GBM width and nonpodocyte cell number per glomerulus, higher mesangial fractional volume, and fewer endothelial fenestrations than those with normoalbuminuria. Participants with macroalbuminuria had higher cortical interstitial fractional volume and lower glomerular filtration surface density and total filtration surface per glomerulus than those with either normoalbuminuria or microalbuminuria. Participants with GFR<90 ml/min at baseline had a higher proportion of globally sclerosed glomeruli, higher cortical interstitial fractional volume, and fewer endothelial fenestrations than those with higher GFR. However, participants with hyperfiltration had higher total filtration surface per glomerulus and higher podocyte number per glomerulus than those with lower GFR. Spearman correlations between structural and functional variables are shown in Table 4. Glomerular volume, filtration surface density, total filtration surface, number of podocytes per glomerulus, and percentage of fenestrated endothelium correlated positively with baseline GFR, and percentage of global sclerosis and mesangial fractional volume correlated inversely.

Table 2.

Glomerular structural parameters in 111 Pima Indians with type 2 diabetes according to the levels of urinary albumin excretion

Characteristic	Normoalbuminuria (n=52)	Microalbuminuria (n=36)	Macroalbuminuria (n=23)	P Valuesa
Global glomerular sclerosis, %	9±12	11±13	14±18	0.44, 0.98, 0.50
Mean glomerular volume, ×106 μ3	5.4±1.6	6.3±1.9	7.0±3.0	0.01, 0.47, 0.01
Glomerular basement membrane width, nm	464±92	528±130	646±135	0.03, 0.003, <0.001
Cortical interstitial fractional volume, %	0.29±0.07	0.28±0.06	0.35±0.09	0.69, 0.002, 0.004
Mesangial fractional volume per glomerulus, %	0.16±0.05	0.19±0.06	0.31±0.10	0.03, <0.001, <0.001
Glomerular filtration surface density, μ2/μ3	0.08±0.02	0.08±0.02	0.05±0.02	0.39, <0.001, <0.001
Total filtration surface per glomerulus, ×105 μ2	4.4±1.7	5.0±1.9	3.1±1.5	0.25, <0.001, 0.001
Nonpodocyte no. per glomerulus	3378±1285	4175±1459	5559±2459	0.01, 0.02, <0.001
Podocyte no. per glomerulus	673±228	664±294	629±346	0.29, 0.36, 0.12
Foot process width, nm	462±164	475±82	516±106	0.10, 0.08, 0.01
Podocyte detachment, %	0.78±1.20	0.85±1.26	1.14±1.37	0.91, 0.51, 0.39
Fenestrated endothelium, %	29.6±7.0	26.7±6.4	20.8±7.3	0.05, 0.003, <0.001

Data are means±SDs.

^aNominal P values are in the order normoalbuminuria versus microalbuminuria, microalbuminuria versus macroalbuminuria, and normoalbuminuria versus macroalbuminuria.

Table 3.

Glomerular structural parameters in 111 Pima Indians with type 2 diabetes according to the levels of GFR

Characteristic	GFR<90 ml/min (n=14)	GFR=90–153 ml/min (n=56)	GFR≥154 ml/min (n=41)	P Valuesa
Global glomerular sclerosis, %	26±16	9±11	7±12	<0.001, 0.28, <0.001
Mean glomerular volume, ×106 μ3	5.7±3.2	5.5±2.0	6.8±1.6	0.85, 0.001, 0.22
Glomerular basement membrane width, nm	572±112	514±139	517±131	0.15, 0.90, 0.17
Cortical interstitial fractional volume, %	0.35±0.08	0.29±0.08	0.29±0.07	0.01, 0.66, 0.01
Mesangial fractional volume per glomerulus, %	0.25±0.08	0.20±0.10	0.18±0.07	0.06, 0.28, 0.003
Glomerular filtration surface density, μ2/μ3	0.06±0.03	0.07±0.03	0.08±0.02	0.17, 0.15, 0.02
Total filtration surface per glomerulus, ×105 μ2	3.2±1.2	3.8±1.5	5.5±1.9	0.18, <0.001, <0.001
Nonpodocyte no. per glomerulus	3981±2155	3859±1919	4438±1557	0.84, 0.12, 0.40
Podocyte no. per glomerulus	536±210	624±257	755±296	0.24, 0.02, 0.01
Foot process width, nm	509±119	462±164	490±188	0.16, 0.33, 0.73
Podocyte detachment, %	1.32±1.80	0.78±1.20	0.72±0.92	0.29, 0.49, 0.25
Fenestrated endothelium, %	22.0±5.7	29.6±7.0	27.5±7.6	0.01, 0.97, 0.02

Data are means±SDs.

^aNominal P values are in the order GFR<90 ml/min versus GFR=90–153 ml/min, GFR=90–153 ml/min versus GFR≥154 ml/min, and GFR<90 ml/min versus GFR≥154 ml/min.

Table 4.

Spearman correlations (P values) between clinical and morphometric variables in 111 Pima Indians with type 2 diabetes mellitus

Characteristic	GS, %	VG, μm3	VvInt, %	VvMes, %	SV, μ2/μ3	TFS, μ2	GBM, nm	E + M, No.	Podo, No.	FPW, nm	PD, %	Fen, %
Age, yr	0.40 (<0.001)a	−0.27 (0.004)a	0.24 (0.01)a	−0.01 (0.91)	0.02 (0.85)	−0.14 (0.14)	−0.19 (0.04)a	−0.20 (0.03)a	−0.21 (0.03)a	0.11 (0.26)	0.04 (0.70)	0.14 (0.15)
Diabetes duration, yr	0.31 (0.001)a	−0.11 (0.27)	0.40 (<0.001)a	0.42 (<0.001)a	−0.31 (0.001)a	−0.37 (<0.001)a	0.24 (0.01)a	0.12 (0.19)	−0.21 (0.03)a	0.166 (0.08)	0.13 (0.17)	−0.22 (0.02)a
MAP, mmHg	−0.01 (0.95)	0.21 (0.03)a	0.11 (0.26)	0.22 (0.02)a	−0.13 (0.17)	−0.03 (0.78)	0.24 (0.01)a	0.35 (<0.001)a	0.05 (0.57)	0.25 (0.01)a	0.09 (0.34)	−0.17 (0.07)
BMI, kg/m2	0.05 (0.63)	0.08 (0.39)	−0.18 (0.06)	−0.23 (0.01)a	0.19 (0.05)	0.23 (0.02)a	−0.28 (0.003)a	−0.02 (0.81)	0.01 (0.92)	−0.15 (0.12)	−0.05 (0.59)	−0.01 (0.92)
HbA1c, %	−0.31 (0.001)a	0.26 (0.01)a	0.08 (0.42)	0.27 (0.004)a	−0.16 (0.10)	−0.02 (0.87)	0.34 (<0.001)a	0.29 (0.002)	0.06 (0.53)	0.17 (0.07)	0.07 (0.44)	−0.19 (0.05)
ACR, mg/g	0.07 (0.44)	0.34 (<0.001)a	0.23 (0.01)a	0.54 (<0.001)a	−0.49 (<0.001)a	−0.23 (0.02)a	0.54 (<0.001)a	0.43 (<0.001)a	−0.15 (0.11)	0.28 (0.003)a	0.12 (0.20)	−0.43 (<0.001)a
GFR, ml/min	−0.29 (0.002)a	0.34 (<0.001)a	−0.29 (0.002)a	−0.31 (<0.001)a	0.27 (0.004)a	0.46 (<0.001)a	−0.10 (0.32)	0.11 (0.27)	0.26 (0.01)a	−0.11 (0.25)	−0.07 (0.46)	0.22 (0.02)a

GS, global sclerosis; V_G, mean glomerular volume; VvInt, cortical interstitial fractional volume; VvMes, mesangial fractional volume; S_V, glomerular filtration surface density; TFS, total filtration surface; GBM, glomerular basement membrane width; E + M, endothelial and mesangial cells per glomerulus; Podo, podocytes per glomerulus; FPW, foot process width; PD, podocyte detachment along glomerular basement membrane; Fen, fenestrated endothelial cells; MAP, mean arterial pressure; BMI, body mass index; HbA1c, hemoglobin A1c; ACR, albumin-to-creatinine ratio.

^aP values <0.05.

In Cox models, higher mesangial fractional volume, percentage of global sclerosis, nonpodocyte number per glomerulus, GBM width, mean glomerular volume, and podocyte foot process width as well as lower filtration surface density and fewer endothelial fenestrations were each significantly associated with RFL after multivariable adjustment (Table 5). A glomerulopathy index, reflecting the combined effects of the statistically significant morphometric variables, was strongly associated with RFL after multivariable adjustment (HR, 1.84; 95% confidence interval, 1.38 to 2.46). When GFR slope was modeled as a threshold, none of the morphometric variables were associated with a slope ≥5 ml/min per year (Table 5), but when modeled as a continuous variable, cortical interstitial fractional volume was associated with slope (Supplemental Table 1). Only the glomerulopathy index remained statistically significant when ACR was included as a covariate in the Cox models (HR, 1.43; 95% confidence interval, 1.03 to 1.97). In each model, ACR was highly statistically significant (data not shown). The addition of the glomerulopathy index to the Cox model that included all covariates and ACR increased the c statistic for predicting RFL from 0.779 to 0.785 (P=0.04). The addition of this index also improved the relative integrated discrimination improvement for predicting RFL by 4.8% after 6 years of follow-up.

Table 5.

Hazard ratio (95% confidence interval) per 1-SD increment of each morphometric variable from Cox regression models for ≥40% loss of GFR and a GFR slope ≥5 ml/min per 1.73 m² per year

Morphometric Variable	1 SD	Hazard Ratios (95% Confidence Intervals) for ≥40% Renal Function Loss		Hazard Ratios (95% Confidence Intervals) for GFR slope ≥5 ml/min per year
		Unadjusted	Adjusteda	Unadjusted	Adjusteda
Global glomerular sclerosis, %	14	1.08 (0.83 to 1.41)	1.63 (1.21 to 2.21)	0.95 (0.71 to 1.27)	1.00 (0.73 to 1.38)
Mean glomerular volume, ×106 μ3	2.1	1.32 (1.02 to 1.70)	1.42 (1.02 to 1.96)	1.18 (0.92 to 1.52)	1.06 (0.78 to 1.43)
Glomerular basement membrane width, nm	133	1.53 (1.16 to 2.00)	1.48 (1.05 to 2.08)	1.41 (1.08 to 1.85)	1.07 (0.78 to 1.47)
Cortical interstitial fractional volume, %	8	1.12 (0.84 to 1.50)	1.29 (0.93 to 1.80)	1.24 (0.90 to 1.71)	1.29 (0.91 to 1.84)
Mesangial fractional volume per glomerulus, %	9	1.84 (1.39 to 2.42)	2.27 (1.58 to 3.26)	1.22 (0.92 to 1.62)	0.98 (0.67 to 1.45)
Glomerular filtration surface density, μ2/μ3	0.03	0.67 (0.48 to 0.92)	0.62 (0.41 to 0.94)	0.90 (0.66 to 1.22)	1.11 (0.77 to 1.60)
Total filtration surface per glomerulus, ×105 µ2	1.8	0.91 (0.69 to 1.21)	0.82 (0.54 to 1.25)	1.09 (0.82 to 1.45)	1.26 (0.85 to 1.86)
Nonpodocyte no. per glomerulus	1829	1.46 (1.15 to 1.87)	1.50 (1.10 to 2.05)	1.23 (0.96 to 1.59)	1.04 (0.78 to 1.40)
Podocyte no. per glomerulus	276	0.95 (0.70 to 1.31)	0.82 (0.56 to 1.18)	1.08 (0.78 to 1.49)	1.04 (0.72 to 1.51)
Foot process width, nm	131	1.29 (1.05 to 1.59)	1.28 (1.03 to 1.60)	1.16 (0.92 to 1.46)	1.08 (0.83 to 1.40)
Podocyte detachment, %	1.25	0.86 (0.63 to 1.16)	0.75 (0.54 to 1.04)	0.88 (0.63 to 1.22)	0.86 (0.61 to 1.20)
Endothelial fenestration, %	7.6	0.68 (0.50 to 0.91)	0.68 (0.48 to 0.95)	0.91 (0.67 to 1.23)	1.03 (0.73 to 1.45)
Glomerulopathy indexb	NA	1.58 (1.22 to 2.06)	1.84 (1.38 to 2.46)	1.30 (1.02 to 1.65)	1.12 (0.84 to 1.50)

Fifty-one events occurred for the first outcome, and 49 occurred for the second outcome. NA, not applicable.

^aAdjusted for baseline age, sex, diabetes duration, hemoglobin A1c, GFR, and treatment assignment during the clinical trial.

^bStandardized sum of the standardized values of morphometric variables significantly associated with renal function loss ≥40% after multivariable adjustment.

In sensitivity analyses, podocyte foot process width and GBM width lost statistical significance when GFR slope from the initiation of the clinical trial was added to the model instead of GFR at the time of biopsy. All other predictors were unchanged. One hundred three participants (93%) reported taking RAS inhibitors after completing the clinical trial. Adjustment of GFR to account for the acute effects of starting or stopping RAS inhibition during follow-up reduced the number of patients with RFL by one and delayed development of RFL in six patients but did not change the morphometric predictors compared with the model that did not adjust for the acute effect of treatment (Supplemental Table 2). When the definition of RFL was modified to require both a ≥40% loss of GFR and a final GFR of <90 ml/min, podocyte foot process width and percentage of endothelial fenestrations lost statistical significance; relationships with all other structural variables were unchanged. When the criteria for RFL were modified to include ≥30% loss of GFR instead of ≥40% loss, the number of progressors increased to 66, cortical interstitial fractional volume became predictive, and GBM width, podocyte foot process width, and percentage of endothelial fenestrations were no longer predictive of RFL. Conversely, when the criteria for RFL were modified to include ≥57% loss of GFR, there were only 27 progressors, and only percentage of global sclerosis predicted RFL (Supplemental Table 3).

Discussion

The classic glomerular lesions of diabetic kidney disease predict RFL in Pima Indians with type 2 diabetes. Major predictors included mesangial expansion; increased percentage of global glomerular sclerosis, mean glomerular volume, and podocyte foot process width; reduced glomerular filtration surface density; and fewer endothelial fenestrations. A glomerulopathy index developed from these morphometric variables was also strongly associated with RFL. These associations were present, even when the baseline GFR was normal or elevated. The higher number of morphometric predictors identified when the RFL outcome was defined as ≥30% GFR loss and the lower number of predictors identified when the outcome was defined as ≥57% GFR loss are likely attributable to the differing numbers of events associated with each definition. When we added ACR to the models, the predictive value of the individual structural variables was lost, but the glomerulopathy index remained statistically significant, modestly enhancing the discrimination of the survival model for the RFL outcome beyond that achievable by traditional risk factors, including ACR. The predictive value of ACR for RFL in this cohort of Pima Indians with type 2 diabetes is likely mediated by specific diabetic glomerular structural lesions. These findings are entirely consistent with recent cross–sectional observations in type 1 diabetes (34).

Podocyte foot process width predicted loss of renal function, and podocyte cell number per glomerulus was modestly lower in progressors versus nonprogressors. The number of nonpodocyte cells per glomerulus was substantially higher than the number of podocytes per glomerulus, consistent with other reports in patients with diabetes (35), and may reflect their contribution to mesangial expansion (36) and thus, reduced glomerular filtration surface (37).

We also reported the decline in GFR as a linear slope and a threshold defined by a linear slope, although we recognize that, by imposing a linear function on a commonly nonlinear disease course, distinct patterns of GFR decline that are more predictive of progressive diabetic kidney disease than others may be obscured (38). Only cortical interstitial fractional volume was associated with linear slope when modeled as a continuous variable, and none of the morphometric variables were associated with linear slope when modeled as a threshold, suggesting that RFL is a better predictor of clinical outcome than linear GFR slope. Indeed, the predictive value of RFL for the major health outcomes of ESRD and death has already been established (17).

The structural-functional relationships illustrated in this study and previously in a subset of this cohort (19) extend into the hyperfiltration range, which is consistent with findings in type 1 diabetes (37,39). Nevertheless, few studies have examined the effect of renal lesions on renal functional changes in type 2 diabetes. In a study involving protocol kidney biopsies in 108 Europeans with type 2 diabetes and elevated urinary albumin excretion, GBM width and mesangial fractional volume predicted GFR decline over 4 years as measured every 6 months by the plasma clearance of chromium-51 EDTA (10). A subset of the participants had a rapid decline in GFR, but they had albumin excretion rates at baseline similar to those with stable kidney function, so that the kidney biopsy offered important prognostic information not provided by the level of albuminuria (10). These findings were confirmed in a cohort of Japanese patients with type 2 diabetes who were followed for 11 years after a protocol kidney biopsy (9).

RAS inhibition affects GFR both acutely and chronically (40,41). The acute effect lowers GFR during the first 1–3 months of treatment; the chronic effect slows the rate of GFR decline, and the magnitude of this effect increases with increasing duration of treatment (41). Losartan may also have epigenetic effects that last well beyond the treatment period (42). Because the acute and chronic effects on GFR are different, accounting for these effects is difficult, particularly when change in GFR is the outcome, such as in this study. In this setting, a commonly used approach of fitting the regression model with RAS treatment as a binary variable is not valid (43). A preferred approach is to adjust the observed GFR in each participant according to changes in RAS treatment during follow-up (43). We reported previously that initiation of RAS treatment was associated with a GFR reduction after 1 month of treatment of 3.75% relative to placebo (20), and therefore, we applied this adjustment to the measured GFR in this study. One less person reached the RFL outcome after adjustment, and the conclusions of the study were unchanged by this additional adjustment.

Strengths of this study include its longitudinal design, detailed characterization of the study cohort, and serial measurement of GFR. Limitations include the use of an arbitrarily defined dichotomous definition of RFL as an outcome for all levels of baseline GFR. Given the higher mean GFR at baseline in the progressors than in the nonprogressors, a ≥40% decline could reflect a substantial fall in which the GFR still remains within the normal or hyperfiltering range. Among participants with extreme initial GFRs, regression to the mean may be a factor in declining GFR. Nevertheless, when we restricted the definition of RFL to those who had a ≥40% decline and also achieved a final GFR <90 ml/min, the results of the study were largely unchanged. Another limitation is that data on medicine usage during the observational study are self-reported, not subject to verification by pill counting, and only collected at yearly intervals. Although we could adjust for the acute effects of RAS inhibition on GFR, we could not adjust for the chronic effects. Virtually all study participants, however, received RAS inhibitors during the observational part of the study, and previous data in this cohort illustrate that the effect of RAS inhibition on GFR, at least in early diabetic kidney disease, is modest (20). Nonetheless, because the vast majority of participants in both treatment arms in the initial trial were prescribed RAS inhibitors during follow-up, this study suggests that timing of initiation of RAS inhibition could have an important influence on renal outcomes.

Our findings show that quantitative measures of glomerular structure predict loss of renal function in type 2 diabetes. Although loss of renal function, when the GFR is still normal or elevated, may also reflect improved glycemic management, reduced protein intake, or other factors, such as aging, a substantial loss, even when GFR remains within a normal range, suggests the presence and progression of underlying renal pathology.

Disclosures

None.